Cooperative Learning

Review of Educational Research http://rer.aera.net

Cooperative Learning Robert E. Slavin REVIEW OF EDUCATIONAL RESEARCH 1980 50: 315 DOI: 10.3102/00346543050002315 The online version of this article can be found at: http://rer.sagepub.com/content/50/2/315

Published on behalf of

American Educational Research Association

and http://www.sagepublications.com

Additional services and information for Review of Educational Research can be found at: Email Alerts: http://rer.aera.net/alerts Subscriptions: http://rer.aera.net/subscriptions Reprints: http://www.aera.net/reprints Permissions: http://www.aera.net/permissions Citations: http://rer.sagepub.com/content/50/2/315.refs.html

Downloaded from http://rer.aera.net at Peking University Library on July 28, 2011

Review of Educational Research Summer, 1980, Vol. 50, No. 2, Pp. 315-342

Cooperative Learning Robert E. Slavin Johns Hopkins

University

Research on classroom cooperative learning techniques, in which students work in small groups and receive rewards or recognition based on their group performance, has been increasing in the pastfew years. This review summarizes the results of 28 primary field projects lasting at least 2 weeks, in which cooperative learning methods were used in elementary or secondary classrooms. The pattern of research findings supports the utility of cooperative learning methods in general for increasing student achievement, positive race relations in desegregated schools, mutual concern among students, student self-esteem, and other positive outcomes. The various cooperative learning methods are contrasted in terms of characteristics and outcomes, and the next steps for research in this area are outlined.

Cooperative learning is an old idea in education, which has experienced a substantial revival in educational research and practice in the past few years. The term refers to classroom techniques in which students work on learning activities in small groups and receive rewards or recognition based on their group's performance. Laboratory research on the effects of cooperation on performance and other variables was already under way in the 1920's (see Mailer, 1929), but only recently have the principles of cooperation been made into practical programs for use in schools and evaluated as such. This paper presents a general theory and typology of cooperative learning, reviews the research on practical cooperative learning programs, and draws implications from this research for a new model of classroom instruction based on cooperation as a dominant instructional mode.

Cooperative Learning and Classroom Technology Classroom instructional technology can be described as a combination of three essential elements: a task structure, a reward structure, and an authority structure. The task structure is the mix of activities that make up the school day. Lecture, class discussion, and seat work are different task structures in use in most classrooms. Another dimension on which task structures may vary is the grouping system in use. Students may be working on individually prescribed tasks, in homogeneous or This work was supported by a grant from the National Institute of Education, No. NIE-6-780210. The opinions expressed herein do not necessarily reflect the position or policy of the National Institute of Education, and no official endorsement by that office should be inferred. The author would like to thank the following individuals for their comments on this paper: Nancy Karweit, Nancy Madden, Stephen Hansell, Carole Ames, and David DeVries.

315 Downloaded from http://rer.aera.net at Peking University Library on July 28, 2011

ROBERT E. SLAVIN

heterogeneous small groups in which students may or may not be permitted to help one another with their work, with or without a teacher or aide, and so on. The reward structure of the classroom may also vary on several dimensions. Rewards for appropriate behavior vary in kind. Some possible rewards include grades, teacher approval, and tangible rewards. They may vary in frequency, in magnitude, and in sensitivity, the degree to which increases in performance are matched with increases in reward.- The term interpersonal reward structure refers to the consequences for an individual of his classmates' performance. In a competitive reward structure, such as grading on the curve, one student's success necessitates another's relative failure. Michaels (1977) calls this negative reward interdependence, because students' rewards are linked to one another negatively. The opposite of competition is cooperation, such as is present in sports teams. In cooperation, or positive reward interdependence, one student's success helps another to be successful. The third interpersonal reward structure is reward independence, or individualization, where students' goal achievements are unrelated to each other. The authority structure of the classroom refers to the control that students exercise over their own activities, as opposed to that exercised by teachers and other adults. In some classrooms students have considerable choice about what they will study, how they will learn what they need to learn, in what order they will do a prescribed set of tasks, and in some cases students have a say in how much they will do to earn a certain grade. The authority structure is relatively unidimensional, varying from high student autonomy in any of several domains of student activities to high teacheror school-imposed structure. Cooperative learning may involve changes in all three of the major elements of classroom technology, but it is primarily a change in the interpersonal reward structure of the classroom, from a competitive reward structure to a cooperative one. Of course, other changes are almost unavoidable, such as a change from a primarily individual task structure with frequent whole-class instruction to a task structure characterized by student interaction in small groups. Other changes in the reward, task, and authority structures of the classroom always accompany the implementation of cooperative learning techniques, and these changes are confounded with the change in reward structure. However, cooperative learning grew out of a laboratory tradition in social psychology that was clearly focused on the changes in reward structures, and the research on cooperative learning involves task and authority structure changes as secondary in interest to the changes in the interpersonal reward structure. A Brief Theory of Reward Structures Although an elaborate theory of reward structures is beyond the scope of this review, a brief statement is necessary for understanding the research. There are two primary outcomes that are important in research on reward structures: performance and cohesiveness. Performance refers to individual and group productivity on any of a variety of tasks; cohesiveness includes such variables as liking of others, feeling of being liked, group evaluation, race relations, and so on. These major categories of outcomes have very different theoretical bases, and are thus discussed separately. Performance. A theory of the effects of different reward structures on performance is anything but straightforward. This is true because changes in reward structures 316


COOPERATIVE LEARNING

often lead to mediating variables that have contradictory effects on performance. The kind of performance and its measure, the task, the particular form of the reward structure, and other factors make these mediating variables more or less important, and thus determine the net direction of the effect on performance. There are several issues that need to be considered in predicting the effects of different reward structures on performance. Some of these are listed as follows: 1. How contingent are an individual's rewards on performance? That is, what are the chances that if an individual works hard, the extra work will be detected, recognized, and rewarded, and that if the individual does not work up to capacity that the rewards will be diminished. This is the probability of success given greater or lesser effort discussed by Slavin (1977a). 2. How large are the potential rewards for performance? Note that with the probability of success described above, this is the familiar incentive value of success— probability of success model proposed by Atkinson (1958) and many others. This model posits a multiplicative relationship between incentive value of success and probability of success. 3. Are others likely to help or hinder task performance? Consideration of the effects of changes in reward structures makes it clear why the answer to the question "Which reward structure has the most positive effects on performance?" must be that it depends on the task, the performance and its measure, and many other features. First, consider cooperation, or positive reward interdependence. In general, participation in a cooperative group reduces the degree to which individual performances lead to individual outcomes. A group member can have an off day and still be successful because his or her groupmates cover for the individual; a group member may work especially hard and still fail because the rest of the group does not match his or her efforts. The larger the group, the greater the reduction in the degree to which rewards depend on an individual's performance (see Slavin, 1977a, for a discussion of this issue). On the other hand, the reward delivered to the group is not the only one present in a cooperative contingency, and it may be the least important one. When individuals in a cooperative contingency want to be rewarded, they have two primary means of increasing their chances. First, they can work hard themselves. But, second, they can try to influence or help their groupmates to do their best. That is, a cooperative contingency sets up a situation in which group members administer a highly contingent reward structure with their groupmates; if they do what helps the group to be rewarded, they receive praise; if they do not, they receive blame. Even if praise and blame are not directly delivered, a cooperative reward structure can create a general group norm favoring performance (Thomas, 1957). Also, group members are motivated to help one another to be successful, and will facilitate one another's performance with whatever means they have. Thus, cooperative reward structures reduce performance by reducing the connection between performance and outcome, and they increase performance by introducing interpersonal rewards for individual behavior, group norms favoring performance that helps the group to achieve its goals, and help among group members. The net outcome depends on how much each of these factors is made important. For example, on some tasks, such as solving a single problem as a group, help from group members may be very useful, so a cooperative reward structure is likely to be effective. If the

317


ROBERT E. SLAVIN

task requires that each group member solve his or her own problem, group help doesn't help so much, and an individualistic or competitive reward structure might be just as good or better. On some tasks, group members can see how their groupmates are doing and can reward them effectively; on other tasks they cannot, so it may be unlikely that group member praise will work to increase individual performance (see Jones & Vroom, 1964). Note that there is very little theoretical distinction between pure group cooperation and competition between groups, in most cases. Looking within the group, group competition is just another form of positive reward interdependence. All of the problems and benefits associated with cooperation as just described apply equally well to group competition; the group competition is just one way of establishing a criterion for group success. The one major distinction between group cooperation and group competition is the possibility of negative affect and hindering between groups, as documented by Sherif and Sherif (1953), but the essential motivational dynamics within the team are the same in either reward structure. The few studies that have compared "pure" group cooperation to group competition have found few differences in effects (e.g. Grossack, 1954; Hammond & Goldman, 1961; Julian & Perry, 1967). In a competitive reward structure, the relationship between performance and outcome depends on the competitive standard, the way rewards are distributed, and the individual's ability (or perceived ability) compared to the performance level needed to be successful. For example, there is no relationship between performance and outcome, and thus very little motivation, in a competition where the competitors are so poorly matched that some cannot win and others cannot lose, or where there is such a large chance component that incremental performance does not result in an increased chance of success. On the other hand, to the degree that competitors are evenly matched and success objectively depends on effort, competition can be a highly contingent reward structure. The size of rewards for competition also depends on the particular task. In addition to extrinsic rewards for competitive success, it is apparently satisfying to many just to "best" a competitor. The degree to which "beating" itself is valued is at least in part culturally determined; white American children, for example, will choose to sacrifice rewards for themselves if they can make their opponents lose even to a greater extent than will Mexican-American or black children (Madsen and Shapiro, 1970). Another variable in the achievement of competitive rewards is the perceived difficulty of the task. As Atkinson (1958) points out, competitive success in a situation with a high probability of success is less satisfying than with a moderate or low probability of success. One set of rewards works against high performance in competitive reward structures. This is the interpersonal reward system. While peer norms and sanctions favor high performance in cooperative settings, they oppose it in competitive ones, because one person's success necessitates another person's failure. This peer opposition to performance is very important in places like schools, where standing out as an academic achiever may lead to social disapproval (Coleman, 1961), and delinquent groups (Graubard, 1969). Similarly, fellow competitors will be motivated to hinder one another's performances. The degree to which this interferes with performance depends on how much hindering hinders. For example, if a group of detectives are competing for a promotion, they might refuse to share clues on a common case that 318



would help all of them solve a crime. On the other hand, if the detectives were working on different cases, it would be harder for their competitors to hinder their performance. Individualization is the simplest reward structure. The connection between performance and outcome depends entirely on how sensitive the reward system is to individual performance variations, how attainable standards are, and so on. Of course, these considerations are also important in cooperative and competitive reward structures. Individuals in an independent reward structure are not motivated to help or hinder one another, or to praise or discourage one another's performance. Cohesiveness. A theory of effects of different reward structures on cohesivenessrelated variables is easier to describe than is the theory of effects on performance, but still has its complications. Cooperation should increase group cohesiveness both because it increases contact among group members (Lott & Lott, 1965) and because people tend to like those who facilitate their own rewards (Johnson & Johnson, 1972). Individualization should not help group cohesiveness, because it does not involve interpersonal contact or facilitation of reward among group members. The case of competition is more complicated. In face-to-face competition, interpersonal contact can be quite high and can be rewarding to the competitors, particularly if they are balanced in ability. Operating in the opposite direction, however, is the fact that competitors frustrate each others' goal achievement, and thus will not be viewed positively. While cooperation should lead to greater group cohesiveness than individualization, the relative effects of competition should depend on the particular form of the competition, including the importance of success to each competitor. Classroom Cooperative Learning

Techniques

The bulk of the research on practical cooperative learning techniques has focused on four major models: Teams-Games-Tournament (DeVries & Slavin, 1978b); Student Teams-Achievement Divisions (Slavin, 1978b); Jigsaw (Aronson, 1978); and Small-Group Teaching (Sharan & Sharan, 1976). These techniques are emphasized in this review both because they have been well researched in field settings and because they are well-defined teaching strategies that are in use in many classrooms. All four have books or manuals written about them so that teachers can easily implement them. Other classroom research involving less widely used cooperative techniques is also reviewed. Teams-Games-Tournament. Teams-Games-Tournament (TGT) is built around two major components: 4- to 5-member student teams, and instructional tournaments. The teams are the cooperative element of TGT. Students are assigned to teams according to a procedure that maximizes heterogeneity of ability levels, sex, and race. The primary function of the team is to prepare its members to do well in the tournament. Following an initial class presentation by the teacher, the teams are given worksheets covering academic material similar to that to be included in the tournament. Teammates study together and quiz each other to be sure that all team members are prepared. After the team practice session, team members must demonstrate their learning in the tournament, which is usually held once each week. For the tournament, students are assigned to three person "tournament tables." The assignment is done so that competition at each table will be fair—the highest three students in past performance 319


ROBERT E. SLAVIN

are assigned to Table 1, the next three to Table 2, and so on. At the tables, the students compete at simple academic games covering content that has been presented in class by the teacher and on the worksheets. Students at the tournament tables are competing as representatives of their teams, and the score each student earns at his or her tournament table is added into an overall team score. Because students are assigned to ability-homogeneous tournament tables, each student has an equal chance of contributing a maximum score to his or her team, as the first place scorer at every table brings the same number of points to his or her team. Following the tournament, the teacher prepares a newsletter which recognizes successful teams and first place scorers. While team assignments always remain the same, tournament table assignments are changed for every tournament according to a system that maintains equality of past performance at each table. For a complete description of Teams-Games-Toumament, see Slavin (1978c). Student Teams-Achievement Divisions. Student Teams-Achievement Divisions (STAD) uses the same 4- to 5-member heterogeneous teams used in TGT, but replaces the games and tournaments with simple, 15-minute quizzes, which students take after studying in their teams. The quiz scores are translated into team scores using a system called "achievement divisions." The quiz scores of the highest six students in past performance are compared, and the top scorer in this group (the achievement division) earns eight points for his or her team, the second scorer earns six points, and so forth. Then the quiz scores of the next highest six students in past performance are compared, and so on. In this way, student scores are compared only with those of an ability-homogeneous reference group instead of the entire class. A "bumping" procedure changes division assignments from week to week to maintain equality. Students know only their own division assignments; they do not interact in any way with the other members of their division. The achievement division feature maintains the equality of opportunity for contributions to the team score as in TGT. A complete description of STAD appears in Slavin (1978c). Jigsaw. In Jigsaw, students are assigned to small heterogeneous teams, as in TGT and STAD. Academic material is broken into as many sections as there are team members. For example, a biography might be broken into "early years," "schooling," "first accomplishments," and so forth. The students study their sections with members of other teams who have the same sections. Then they return to their teams and teach their sections to the other team members. Finally, all team members are quizzed on the entire unit. The quiz scores contribute to individual grades, not to a team score as in TGT and STAD. In this sense, the Jigsaw technique may be seen as high in task interdependence but low in reward interdependence, as individual performances do not contribute directly to a group goal. In the Jigsaw technique, individual performances contribute to others' individual goals only; since the group is not rewarded as a group, there is no formal group goal. However, because the positive behavior of each team member (learning the sections) helps the other group members to be rewarded (because they need each other's information), the essential dynamics of the cooperative reward structure are present. Slavin (1978a) constructed a modification of Jigsaw called Jigsaw II. In Jigsaw II, students all read the same material but focus on separate topics. The students from different teams who have the same topics meet to discuss their topics, and then return to teach them to their teammates. The team members then take a quiz, and the quiz

320



scores are used to form team scores as in STAD. Thus, Jigsaw II involves less task interdependence and more reward interdependence than Jigsaw. Small-Group Teaching. Small-Group Teaching is a general classroom organizational plan in which learning takes place through cooperative group inquiry, discussion, and data gathering by students. Students select subtopics within a general area selected by the teacher, and then organize themselves into small groups of two to six members. These groups further subdivide their topic into individual tasks to be performed by group members in preparation for a group presentation to the total class. This group presentation is then evaluated by the other students and by the teacher. Thus, Small-Group Teaching is very high in student autonomy and involves a high degree of task interdependence because of the assignment of students to special tasks within the group, but it is relatively low in group reward interdependence (group rewards are not well-defined) and individual accountability. Other Classroom

Studies

In addition to the four major techniques, there have been several studies of cooperative learning that contribute to understanding of their effects. They differ from the four techniques just described primarily in that they have been used almost exclusively in research, and are not in wide classroom use. Three of these studies (Johnson, Johnson, Johnson, & Anderson, 1976; Johnson, Johnson, & Scott, 1978; Cooper, Johnson, Johnson, & Wilderson, in press) used a simple cooperative technique in which students were assigned to small groups and instructed to work together on academic tasks and hand in a single assignment as a group. The teachers praised the group as a whole, but no formal group rewards were given. This cooperative technique resembles practical techniques presented in Johnson and Johnson's book, Learning Together and Alone (1975), but are not similar enough for the studies of them to constitute an evaluation of the strategies outlined in the book. Two studies (Wheeler & Ryan, 1973; Wheeler, Note 1) used a cooperative technique like that used by the Johnsons, but considerably more structured. Students were assigned roles within cooperative groups and worked on social studies inquiry activities to produce a single workbook. The group making the best workbook received a prize. Weigel, Wiser, and Cook (1975) used a combination of cooperative techniques over a long period of time (in their junior high school sample an entire school year). These techniques involved various small-group activities, with information-gathering, discussion, and interpretation conducted by the student groups. Prizes were given to winning groups based on the quality of the group product. Hamblin, Hathaway, and Wodarski (1971) used a group contingency for academic performance, in which students earned rewards based on (1) the lowest three quiz scores in the group, (2) the highest three scores, or (3) the average group score. Students were encouraged to work together to improve their scores. A Typology of Cooperative Learning Techniques Classroom cooperative learning techniques differ primarily along five dimensions: reward interdependence, task interdependence, individual accountability, teacherimposed structure, and use or nonuse of group competition. High reward interde321


ROBERT E. SLAVIN

pendence means that there is an explicit group reward based on the group's performance. Low reward interdependence describes a situation in which students are asked to work with one another and are praised as a group, but group performance does not lead to a concrete goal in any systematic way. Jigsaw represents a special case of low reward interdependence; there is no formal group goal, but the task interdependence is so extreme that reward interdependence is indirectly created. Students cannot do well on their quizzes unless their teammates teach them well, as each group member has unique information. High task interdependence refers to a situation in which students must rely on one another to do their group tasks. In low task interdependence, individual students could opt to work alone without disrupting the group activity. Of course, "high" and "low" task interdependence are relative terms among cooperative learning techniques; even a technique very low in task interdependence would be high in comparison to a traditional, individual task structure. High individual accountability means that team members' contributions to their team scores are separately quantifiable. For example, in TGT and STAD the team scores are made up of the sum of individual, quantifiable scores. This is in contrast to the Johnson techniques, where a single paper is handed in by the group and individual contributions are impossible to quantify. Individual accountability is a particularly important feature, as without it, it is possible for group members to let others do most of the work in meeting the group goal. The opposite of individual accountability is substitutability, the condition in which all group members have the same task and can substitute for one another in performing the task. Teacherimposed structure refers to the degree to which tasks, rewards, and schedules are imposed by the teacher (or by the technique). The opposite of teacher-imposed structure is high student autonomy and student participation in classroom decisionmaking. The use of group competition means that a prize or recognition is given to the highest scoring groups in the class. Table I summarizes the cooperative techniques including Jigsaw II in terms of reward interdependence, task interdependence, individual accountability, teacherimposed structure, and use of group competition. As is indicated in Table I, the techniques vary a great deal along the five dimensions. At present there are no data to inform a prediction about the separate effects of these dimensions on any of the dependent variables, as they have not been systematically varied. Most effects of cooperative reward structures have been attributed to reward interdependence (see, for example, Johnson & Johnson, 1974), yet at least some degree of task interdependence is usually confounded with the reward interdependence. The typology presented in Table I gives a rough outline of the characteristics of cooperative learning techniques that is useful in understanding the different effects that these methods have had on major student outcomes. The Research The research on the various cooperative learning techniques has been conducted in field experiments, in which the cooperative techniques were compared to control classes and often (in the TGT and STAD studies) to modifications or components of the techniques. This review is restricted to field studies in which practical techniques



were implemented by teachers in classrooms for at least two weeks to differentiate them from shorter term laboratory studies. The studies vary widely in terms of designs, measures, populations, and other features, which makes it difficult to compare results from study to study or, more importantly, from technique to technique. For example, the Johnson studies and some of the TGT studies used random assignment of students to treatments, while in the STAD studies and the other TGT studies, random assignment was done only at the classroom level. In the Jigsaw and Small-Group Teaching studies, teachers who volunteered to use the cooperative techniques were "matched" with control teachers who did not have a chance to volunteer for the experimental groups. In most of the STAD and Jigsaw studies the control groups did exactly what the experimental groups did, including studying the same curriculum materials and following the same schedule of testing and worksheet work. The control groups were given more general objectives and left to their own devices to achieve them in most of the TGT and Small-Group Teaching studies. There are additional differences in measures, statistical procedures (including use or nonuse of pretests and other information as covariates), and other features that make direct comparison of techniques by looking at presence or absence of positive results in various studies difficult. Table II summarizes the major features of the research on cooperative learning techniques. Outcomes Three outcomes of cooperative learning are emphasized in this review: academic achievement, race relations, and mutual concern among students. These were chosen because they are the outcomes most common in studies of cooperative learning that are most valued by schools. Other outcomes that have also been measured in cooperative learning studies but are not presented in Table III include student liking of school, self-esteem, time on-task, ability to take the perspective of another person, and various measures of cooperativeness and competitiveness. These outcomes are also very important, but they were not measured frequently enough in the different studies for comparisons to be made between techniques.

TABLE I Characteristics of Cooperative Learning Techniques Techniques

Reward Interdependence

Task Interdependence

Individual Accountability

TeacherImposed Structure

Group Competition Used

TGT STAD Jigsaw Jigsaw II Small-Group Teaching Johnson techniques Wheeler techniques Weigel et al. techniques Hamblin et al. techniques

High High Low High Low Low High High High

Low Low High High High Low High Low Low

High High High High Low Low Low Low High

High High High High Low Low High Low High

Yes Yes No Yes No No Yes Yes No

323


R O B E R T E.

SLAVIN

TABLE II Characteristics of Cooperative Learning Studies Number of Stu dents

Major Reports

Grade Level

Duration (Weeks)

Level of Random Assignment

Kind of School

TGT 1. Edwards, DeVries, & Snyder(l972) 2. DeVries & Edwards (1973) 3. Edwards & DeVries (Note 4) 4. Hulten & DeVries (Note 5) 5. DeVries, Edwards, & Wells (Note 6) 6. DeVries & Mescon (Note 7) 7. DeVries, Mescon, & Shackman (Note 8) 8. DeVries, Mescon, & Shackman (Note 9) 9. DeVries, Lucasse, & Shackman (Note 10) 10. Slavin, (1977c)

96

1

9

Class

Urban East

110

1

4

Student

Urban East

128

Ί

12

Student

Urban East

299

Ί

10

Class

Urban East

191

10-12

12

Class

Suburban South

60

3

6

Student

Suburban East

53

3

6

Student

Suburban East

53

3

5

Student

Suburban East

1742

7-8

10

Teacher

57

7-9

10

Teacher

Suburban Mid west Suburban East (Emotionally Disturbed Ad olescents)

STAD 11. Slavin, (1978a)

205

Ί

10

Class

12. 13. 14. 15.

Slavin, (1977b) Slavin (in press) Slavin (1979) Slavin & Oickle (Note 11) 16. Madden & Slavin (Note 12)

62 424 424 230

Ί 4 7-8 6-8

10 12 12 12

Class Class Class Class

Eastern Rural Town Urban East Rural East Urban East Rural East

175

3-6

6

Class

Urban East

17. Blaney, Stephan, Rosenfíeld, Aronson, &Sikes(1977) 18. Lucker, Rosenfīeld, Sikes, & Aronson (1976) 19. Gonzales (Note 3)

304

5-6

6

Non-Ran dom

Urban Southwest

303

5-6

2

Non-Ran dom

Urban Southwest

326

9-12

10

Norn-Ran dom

Western Rural Town

Jigsaw

Combined Program (TGT, STAD, Jigsaw II) 20. Slavin (Note 13)

559

4-5

16

Non-Random

324


Rural East

COOPERATIVE LEARNING T A B L E II Number of Students

Major Reports

Grade Level

Continued Duration (Weeks)

Level of Random Assignment

Kind of School

Small-Group Teaching 21. Sharan, Hertz-Lazarowitz, & Ackerman (in press)

217

22. Johnson et al. (1976) 23. Johnson, Johnson, & Scott (1978) 24. Cooper et al. (in press) 25. Wheeler & Ryan (1973)

30 30

5 5-6

60 88

26. Wheeler (Note 1)

2-6

3

Non-Random

Israel

4 10

Student Student

7 5-6

3 4

Student Student

40

5-6

2

Student

27. Weigel, Wiser, & Cook (1975)

324

7-10

Class

28. Hamblin, Hathaway, & Wodarski(l97l)

38

4

30 (Gr. 7) 20 (Gr. 10) 3

Urban Midwest Suburban Midwest Urban Midwest Suburban Midwest Southern Rural Town Urban West

Student

Urban Midwest

Other Studies

The measurement of "achievement" is always difficult. Many studies used curriculum-specific tests to measure achievement. This is legitimate only when the content and rate of progress in experimental and control groups is held strictly constant; otherwise, students could be learning things other than what they were tested on to different degrees in the different treatments. Other studies used only standardized tests, which have the problem of being so comprehensive that they are insensitive to change due to treatments. Achievement tests can measure facts and low-level skills, or they may measure high-level understanding of material. Only the Sharan studies make a specific breakdown of treatment effects by the concept level of the test, but lack of concrete understanding of what constitutes "achievement" may explain contradictory outcomes in different studies. The same may be true of race relations, mutual concern, and the other variables. Some researchers used sociometric measures of cross-racial friendship choices with unlimited choices, others used limited choices, others used attitude scales without sociometric measures. Mutual concern was measured by questionnaires, by sociometric devices, or by interview. Again, differences in measures rather than differences in actual program effects may account for apparent program differences. Table III summarizes the effects of the various cooperative learning techniques on academic achievement, race relations, and mutual concern among students. A " + " indicates a statistically significant effect as compared to a control group; a "0" indicates no difference, and a " - " indicates a negative effect. In many cases the existence of a " + " is qualified in the text by the observation that positive effects were

325


R O B E R T E.

SLAVIN

TABLE III Effects of Cooperative Learning Techniques on Learning, Race Relations, and Mutual Concern Academic Achievement Study Number

Subject Area

Curriculum-Specific Test

Standardized Test

Race Relations

Mutual Concern

TGT 1 2 3a 3b 4 5 6 7 8a 8b 9 10

Mathematics Mathematics Mathematics Social Studies Mathematics Social Studies Language Arts Language Arts Reading Vocabulary Verbal Analogies Language Arts Social Studies

11 12 13 14 15 16

Language Arts Language Arts Language Arts Language Arts Language Arts Mathematics

17 18 19

Social Studies Social Studies Social Studies

+ + +

+

+ +

+ +

+

+

0

0

+ + + + + + 0

+ 0

+ +

0 0

0 0

+ +

STAD

+

0

0

+ +

+ +

+

0 0 +

0

+

+ +

0

+ + 0

Jigsaw 0

+ + Combined Program

20a 20b 20c 2Od

Mathematics Language Arts (STAD) Social Studies (Jigsaw II) Reading (Jigsaw II or STAD)

21 21a 21b 21c 2ld 2le

Social Studies (a) Grade 2 (b) Grade 3 (c) Grade 4 (d) Grade 5 (e) Grade 6

22 23 24

Language Arts Mathematics Science, English, Geography

0

+

+ 0

+

Small-group Teaching

+ 0

+ 0 + Other

Studies 0

-

326


+

+

COOPERATIVE

LEARNING

TABLE III Continued Academic Achievement Study Number

25 26 27 28

Subject Area

Social Studies Social Studies English Spelling, Mathematics, Reading

Curriculum-Specific Test

Standardized Test

Race Relations

Mutual Concern

0

+ + +

Note. + = statistically significant effect favoring experimental group, 0 = no difference, - = statistically significant effect favoring control group. found on some but not all measures, or were only marginally significant. The standardized tests vary depending on the subject matter and the grade levels, while the curriculum-specific tests are tests made up by the experimenters to measure the objectives pursued by the experimental and control groups. The studies listed in Table II and III are all those known to the author that met the criterion of at least 2 week's duration and measured one or more of the three major dependent variables: achievement, race relations, and mutual concern. Teams-Games-Tournament. TGT is the most widely researched of the cooperative learning techniques. As is indicated in Table II, 10 studies varying widely in subject area, design, populations, and other features have been conducted to evaluate TGT. Table III shows that TGT has also been used in the largest number of studies in which positive effects on academic achievement were found on at least one measure (nine). Interestingly, two of the three studies in which the subject matter was social studies failed to find significantly positive effects on the standardized test or on the curriculum-specific test. The third study in social studies (No. 5) found only a marginally significant result (p < .10). Four of the TGT studies took place in integrated schools. In three of those the TGT students named a larger number of their classmates of other races as "friends" on a sociometric instrument than did the control students, controlling for pretest cross-racial choices (DeVries, Edwards, & Slavin, 1978). In two of these studies, the percentage of cross-racial choices made over all choices was within five percentage points of the choice pattern that would have been observed if race were not a criterion for friendship. This is a remarkable finding in consideration of the frequent finding that integrated students decrease over time in cross-racial friendship choices (see Gerard & Miller, 1975), but it is not at all unusual among cooperative learning techniques of all kinds, as this review demonstrates. Mutual concern among students was measured in seven of the TGT studies. The measure used was a questionnaire scale adapted from Walberg and Anderson's (1968) Learning Environment Inventory, which consists of items concerning students' liking of their classmates and feelings of being liked by them. Positive effects on this variable were found in five of the seven studies. The two in which these effects were not found were studies in third grades, where the scales were less reliable than in the higher grades and where there were problems with ceiling effects. Several of the TGT studies found additional positive effects of this strategy. In


ROBERT E. SLAVIN

four of nine studies in which student liking of school was measured, positive effects for TGT were found. The other five showed no difference; again, in most cases this was due to initially positive attitudes toward school in both TGT and control classrooms. In seven studies, peer norms concerning the importance of doing well academically were measured; in five of these, TGT students exceeded control students in the degree to which they felt others wanted them to do their best academically. In one study (No. 9), positive TGT effects were found on the Social Self-Esteem scale of the Coopersmith Self-Esteem Inventory (Coopersmith, 1975). Study No. 10 involved emotionally disturbed adolescents. In that study, positive effects of TGT were found on increasing interpersonal attraction among students; these effects were maintained in a follow-up 5 months after the study. Three of the TGT studies examined variations of TGT to discover which components accounted for the observed effects. Study No. 4 used a 2 × 2 factorial design to determine the separate effects of the team reward and team practice components of TGT. This study found that while the team reward was important in increasing achievement, the team practice (peer tutoring) was not. This unexpected finding was replicated for STAD in study No. 13, and indicates that it is possibly the use of team recognition and the peer norms favoring achievement that result from cooperative reward structures (see Thomas, 1957) which account for the positive effects of teams on performance, not the opportunity for students to interact and help one another. Two of the studies (Nos. 3 and 5) varied the systems used to compute team scores. In study No. 3, the performance of high achievers was weighted by having scores earned by high achievers count more than those earned by low achievers, while in study No. 5 the scores of the low achievers were weighted. Neither of these scoring systems had any effect on student performance or attitudes. Although many of the TGT studies investigated interaction of the treatments with past achievement and sex, no consistent patterns of interaction were observed for either variable. No sex x treatment interactions were found, and while one study (No. 16) found TGT to be more effective with low performers than high performers, another study (No. 2) found the opposite pattern. It seems safest to say that TGT probably has equal effects on students of both sexes and all levels of past achievement. Possible interactions of TGT effects with student age and the subject matter studied are difficult to determine, as the different levels of these variables are difficult to evaluate in the same study. However, the overall pattern of results suggests that TGT is less effective in social studies than in more objective subjects such as mathematics, language arts, and reading vocabulary. There is no clear pattern of results with respect to student age. In summary, TGT has been shown to have relatively consistent positive results on student achievement, race relations, mutual concerns, and other variables. The effects have been obtained in widely varying settings, subject areas, and grades, which argues strongly for the generality and strength of the technique. The research strongly suggests that the effects on achievement appear to be due more to the use of a cooperative reward structure than to the opportunity for peer tutoring provided by the teams, and these effects have few interactions with subject or setting characteristics. Student Teams-Achievement Divisions. STAD is the simplest cooperative learning technique of those reviewed here; in its basic form it is simply a group contingency for academic performance, where "academic performance" is defined as better 328



performance than that of a comparison group (the achievement division). As such, it is of particular research interest as a relatively "pure" form of a cooperative technique, without many of the additional components that make the origin of the effects for the other techniques unclear. As is shown in Table HI, STAD exceeded control techniques in increasing learning in four of the six studies in which achievement was measured. There is no pattern to these results in terms of subject matter (which was constant) or grade level. One possible reason the achievement results for STAD are somewhat less consistent than those for TGT is that the control groups for STAD were more stringent. In each case, control students studied exactly the same academic materials on the same schedule as the STAD students. In all of the TGT studies except No. 10, the control groups had the same objectives and had the same material available as the TGT groups, but the teachers were not held to the same highly structured schedule of instruction. Of course, the use of the games in TGT could also explain the difference in achievement effects. The findings of positive achievement effects even in contrast to the structured control groups gives stronger support to the conclusion that STAD does increase academic achievement. The effects of STAD on race relations are even stronger and more consistent than those for TGT. All three STAD studies in integrated schools showed positive effects of teams on cross-racial friendship choices. Further, a followup of study No. 14 found that 9 months after the conclusion of the study, students who had been in the experimental (STAD) groups still made significantly more cross-racial friendship choices than former control group students. A secondary analysis of study No. 14 also showed that the treatment effects on cross-racial friendships were equal for blacks and whites, boys and girls, and high and low achievers, and that they were stronger for close, mutual friendship choices than distant or unreciprocated ones (Hansell & Slavin, Note 2). Positive effects of STAD on mutual concern were found in three of six studies. Behavioral observers were used in four studies to determine the percentage of time that students spent on-task (focusing on learning activities). In three of these, STAD students were on-task more than control students; in the fourth (No. 15) there was no difference. In one of the studies (No. 16), the students were mainstreamed, academically handicapped as well as normal progress students. In addition to increasing achievement for all students, the STAD treatment reduced the number of mainstreamed students rejected by their normal progress classmates on a sociometric instrument. Positive effects for all students on self esteem were also found in this study. Two of the STAD studies used factorial designs to separately assess important components of the technique. In study No. 13, a 2 × 2 factorial design varied team reward and team practice (peer tutoring), replicating the design of TGT study No. 4. This study similarly found positive effects of the team factor on academic achievement, but not the team practice. This study also used an external, completely untreated control group, which was simply given the study objectives but not given the focused schedule of instruction used in all of the treatments in the factorial design. The performance of this group was much lower than that of any of the treatments that used the focused schedule, supporting a conclusion that the focused schedule of teaching, worksheet work, and quiz is a major explanation for the effectiveness of STAD and probably TGT. 329


ROBERT E. SLAVIN

In study No. 11, a 2 × 2 factorial design varied use or nonuse of student teams and achievement divisions. No effects were found for academic achievement, but for the percent of time on-task the team factor was much more important than the achievement divisions. As was the case for TGT, no consistent interactions *were found between the treatment effects and sex or past achievement. However, in studies Nos. 12 and 15, a substantial race x treatment interaction was found on achievement, due to greater gains relative to controls by blacks than by whites. This interaction did not replicate in one study, No. 14. An interaction between race and treatment for race relations, favoring greater increases in blacks being listed as friends by whites, than whites by blacks in the experimental group, was found in study No. 15, but no such interaction was found in study No. 14 (see Hansell & Slavin, Note 2). In summary, the research on STAD further supports the positive effects of structured cooperative learning techniques on academic achievement and race relations. One study supports the finding of a TGT study (No. 4) that the achievement effects are due to the team reward rather than team practice; a second study found further that a significant part (but not all) of the effect of STAD on learning was due to the use of a focused schedule of instruction. The positive effects of STAD on students' time on-task adds to a suggestion that STAD may impact achievement in part because it is a highly structured approach to classroom instruction. Jigsaw. There have been only two studies of Jigsaw that measured any of the three major variables in Table II, although two additional studies involved other variables. Jigsaw is the cooperative learning technique, which used the most structured form of team interaction. In TGT and STAD students are given a goal of making sure they and their teammates know a set of academic materials so that all can do well in a quiz or game, but the way they do this is not clearly specified. Jigsaw, with its structured cooperation on the group task but no explicit group goal, is an example of low reward interdependence and high task interdependence. Positive effects of Jigsaw on achievement have been found in one study, No. 18. In that study, which lasted only two weeks, black and Chicano students in the Jigsaw groups learned more than those in the control classes, but there were no differences between the Anglos in each treatment. A second study (No. 17) and two pilot studies preceding it found no differences between Jigsaw and control groups in achievement, although these studies were not specifically designed to assess student achievement. The effects of Jigsaw on race relations is also apparently positive, but the evidence is indirect. In study No. 17, which took place in 13 integrated classrooms (10 experimental, 3 control), Jigsaw students gained in liking for their groupmates more than for their nongroupmate classmates. Because assignment to groups was done to maximize heterogeneity on ethnicity, a chooser's groupmates are more likely to be of another ethnic group than the class as a whole; thus, a gain in liking of groupmates could possibly be a gain in cross-ethnic attraction. However, lacking a direct measure of cross-ethnic attraction, this effect must be considered questionable. Gonzales (Note 3) found more positive attitudes toward Mexican-Americans by Anglos as a consequence of a Jigsaw intervention, but found no effects on attitudes toward Anglos or Asians. Jigsaw effects on mutual concern are also unclear. In study No. 17, there were no experimental-control differences in overall liking of classmates, but the control group increased significantly more than the experimental group in feelings of being liked, 330



a most unusual finding in research on cooperative groups. On the other hand, liking among groupmates increased in the Jigsaw classes, while liking for nongroupmates did not decrease, suggesting a positive effect on mutual concern. Again, this set of findings is in need of replication. One finding that is particularly characteristic of Jigsaw is a positive effect on students' self-esteem. This was found in study No. 17 and was replicated by Geffner (Note 14), although Gonzales (Note 3) failed to find the effect. These effects are probably due to Jigsaw's system of giving each group member unique information needed by the group, thus making each student a special resource for his or her group. Study No. 17 also found positive effects for Jigsaw on liking of school for Anglos and blacks, but not Chicanos, and found that while Jigsaw students felt less competitive toward their classmates over time, control students felt more so. Finally, Jigsaw students saw their classmates as learning resources more than did control students. Bridgeman (Note 15) found that participation in Jigsaw increased children's abilities to take the perspective of others, and Gonzales found that Jigsaw increased students' internal locus of control. In summary, Jigsaw has had relatively consistent positive effects on student selfesteem. The effects are unclear for academic achievement, race relations, liking of school, and mutual concern. Combined program. In one study, three of the major cooperative learning methods were combined for use with the same students at the same time. In this study (No. 20), fourth- and fifth-grade students experienced TGT in mathematics, STAD in language arts, and a modification of Jigsaw in social studies. The modification of Jigsaw (Jigsaw II) involved having students read all of a set of narrative material and become "experts" on particular topics, rather than having the material cut into mutually exclusive sections. Further, the quiz scores earned by individual students were summed into team scores, and team scores were recognized in a class newsletter (as in TGT and STAD). Thus, Jigsaw II encompasses both high-reward interde~ pend«nce and high-task interdependence. The experimental intervention lasted for 16 weeks and occupied more than half of the students' instructional time; thus, this study represents a test of the duration and intensity with which these structured cooperative learning techniques can be used. On seven subscales of the Comprehensive Test of Basic Skills, the experimental classes significantly exceeded control classes (controlling for pretests) on three: Language Mechanics, Language Expression, and Reading Vocabulary. The language effects were ascribed to STAD, and the reading vocabulary effects to Jigsaw II (although they could also have been due to STAD). No differences were found for mathematics (TGT) or social studies (Jigsaw II). Positive effects of the cooperative learning treatments were also found on general self-esteem, and (marginally) on academic self-esteem, two subscales of the Coopersmith Self-Esteem Inventory (Coopersmith; 1975). Interestingly, no differences were found on social self-esteem or on the LEI-derived mutual concern scale, but experimental students named substantially more of their classmates as friends than did control students, a measure of mutual concern among students. Experimental students also expressed greater liking of school than did control students. This study assessed one issue that many critics of the use of team competition have raised. Students were asked to list both their friends in school and the students with whom they would not like to work in a group. It was thought that because of the team competition, students


ROBERT E. SLAVIN

might increase both in friends (their groupmates) and in nonfriends (members of other teams), as occurred in Sherif and Sherif s (1953) Robber's Cave study. However, this did not occur; in fact, the experimental students listed significantly fewer classmates (p < .10) with whom they preferred not to work than did the control students. Small-Group Teaching. Only one study (No. 21) has assessed the effects of SmallGroup Teaching on any of the dependent variables catalogued in Table III, and this study measured only achievement effects. One experimental class was compared to one control class in each of the five grades (2 to 6) in an Israeli school. The students studied different social studies topics in each grade, but the same material in the experimental and control classes within grade. Results indicated no differences in learning of low-level information (such as memory, understanding, and description) except for a positive effect for second graders, but found positive effects for SmallGroup Teaching on high-level concepts (such as identifying concepts, analysis of problems, judgment, and evaluation) in grades 2, 4, and 6. One additional study (Hertz-Lazarowitz, Sharan, & Steinberg, Note 16) found that students who had participated in classrooms using Small-Group Teaching were more cooperative, more altruistic, less competitive, and less vengeful than control students on a reward-allocation task. Other classroom studies. Two of the three studies conducted by the Johnsons and their colleagues measured academic achievement. Both measured achievement in two ways: once by allowing the cooperatively taught students to work in their teams on the tests used as the dependent variables, but having the individually or competitively taught students work alone (the "congruent test"), and once by having all students work alone. The congruent test gives an obvious advantage to the cooperative group, as these students can pool their contributions and rely on the better students to pull up the group scores. As a result, only the tests given individually to both groups are reported here. n study No. 22, cooperative and individualistic groups did not differ in achievement. However, in study No. 23, students in an individualized condition learned marginally more than those in the cooperative groups (p < .10); on a retention test 2 months after the experiment, the individualized group's advantage was significant at the .05 level. This is the only case in the literature on classroom learning groups where the cooperative group performed less well than an individualized control group. One additional study assessed the effects of the Johnson cooperative techniques on race relations. In this study (No. 25), students who had worked in cooperative groups made cross-ethnic friendship choices more often than those in competitive groups. A similar pattern was found when students named those with whom they would like to party and with whom they would like to do schoolwork. Students in the cooperative and competitive conditions also chose learning disabled and emotionally disturbed classmates as friends more often than did students in the individualized groups. The finding of greater cross-racial attraction for both the cooperative and the competitive groups as compared to the individualized groups is interesting; it probably has to do with the fact that the competitive students worked in "clusters," while the individualized students worked entirely alone. The positive effect in the "clusters" of competing students was probably due to the increased contact within the clusters, possibly calling into question the importance of the cooperative reward structure (as



contrasted with simply providing for inteφersonal contact) in increasing cross-racial choices. In one study, not listed in the tables, the Johnson group assessed the effects of cooperation on relations among heterogeneous peers. Armstrong, Balow, and John son (Note 17) found that normal-progress students who had been in cooperative groups with learning disabled children evaluated them more positively than did students in individualized classes. Mutual concern was measured in the Johnson studies as ratings of peers and feelings of being liked by peers. Positive effects on mutual concern variables were found in study No. 22 and in study No. 24. Study No. 23 also found that students who had been in cooperative groups thought they were doing better in school than did individualized students, and they felt they were better liked by their teachers. In study No. 22, cooperative students were found to be more altruistic and to feel that they were better liked by their teachers than were individualized students. Wheeler's studies found conflicting evidence concerning social studies achieve ment. In study No. 25 there were no differences between cooperative and competitive conditions in achievement. In study No. 26, the cooperative group had higher achievement than a competitive group, but students who were cooperatively predis posed (as determined by a reward-allocation task) achieved substantially more in the cooperative groups than in competitive ones, while students with a competitive disposition achieved more in the competitive groups. Weigel, Wiser, and Cook (1975) used their cooperative intervention in seventhand tenth-grade classrooms that were composed of black, white, and MexicanAmerican students. They found that the experimental classes engaged in less crossethnic conflict than did control classes, and that whites in the experimental classes had more positive attitudes toward Mexican-Americans than did whites in control classes. However, no effects were found on attitudes between blacks and whites or blacks and Mexican-Americans, or on Mexican-American attitudes toward whites. Assessment of teacher attitudes clearly indicated support for the cooperative project. Hamblin, Hathaway, and Wodarski (1971) found greater reading, spelling, and mathematics achievement in a simple cooperative reward structure in which groups of fourth graders were rewarded based on the lowest three scores in their groups than in an individual performance contingency. However, when the group contingency was based on the scores of the highest three scorers or on the group average, there were no differences. What Have We Learned? The outcomes of cooperative learning techniques fall mainly into two categories: academic achievement and group cohesiveness. The effects of the techniques on the group cohesiveness variables, such as mutual concern and race relations, are un questionably positive. The achievement results, though usually positive, seem to depend on the particular techniques, settings, measures, experimental designs, or other characteristics. This section considers possible explanations for these different effects in different studies. Achievement. Why have some of the cooperative learning techniques and some of the individual studies shown more positive effects on academic achievement than others? One obvious possible explanation for different findings in a program of 333


ROBERT E. SLAVIN

research is differences in methods. More rigorous studies might be less likely to find significantly positive effects than less rigorous ones. Experimental rigor does explain some of the differences between studies (and between techniques). The TGT studies have the largest number of positive effects on achievement of all the cooperative techniques. Many of the studies used control groups in which the teachers were given the experimental groups' objectives and curriculum materials, but were not held to the same schedule of instruction as the experimental groups. STAD, which is similar to TGT, had somewhat less strong effects on achievement, and all of the STAD studies used the parallel schedule of activities in the control group. This might indicate that the special, focused schedule of activities explains part of the effect of TGT (as well as STAD when it is compared to untreated classrooms), and in fact such an argument is made by Slavin (1978a). It is of further interest that in the studies on Jigsaw and the Johnson techniques (but not Small-group Teaching, which gave the teachers the same materials but did not standardize the schedule of instruction) the curriculum was similarly standardized between the experimental and control groups. In the research done to date, Jigsaw and the Johnson techniques have been less successful than TGT and STAD in increasing academic achievement. Other methodological differences between studies do not seem to be related to the outcomes. Studies that did not use random assignment or that conducted random assignment at the class level rather than the student level were not particularly more or less successful than those that were able to randomly assign students to treatments. One feature that may discriminate studies with positive achievement effects from those without is subject matter. Within the TGT studies, effects were much less likely to be found in social studies than in mathematics, language arts, or reading vocabulary. The combined study found effects on language arts and reading vocabulary, but not social studies. The fact that the Jigsaw studies and one of the Johnson studies were in social studies might contribute to the weak effects of these techniques, although one of the Wheeler studies found greater social studies achievement in the cooperative groups than in control groups. The Small-Group Teaching Study, also in social studies, might shed some light on why the effects of cooperative techniques on social studies achievement might be so spotty. Sharan et al. (in press) divided their dependent measures into "high cognitive level" and "low level" sets of items. They found positive effects on high level items in three of the five grade levels in which they conducted their study, but found positive effects on the low level items only in the second grade. The dependent measures in the TGT, Jigsaw, and the Johnson studies, as well as the social studies measures in the combined program, were undifferentiated by conceptual level and were apparently weighted toward the low level skills. If these studies had broken their measures into high and low level items, the achievement effects on the high level items might have been positive. The successful Wheeler (Note 1) study also assessed understanding of a process, not simply factual knowledge. One problem that is particularly serious in social studies is that unless the teaching content is very explicitly structured, it is difficult to make certain that what is taught corresponds to what is measured. Students in cooperative social studies classes may be learning effective ways to approach high level material that will show up on a posttest, but their learning of the facts may not correspond enough to the content of the posttest to show a difference. One relatively consistent difference between the techniques, which is not completely explained by differences in methodology or subject matter, is the stronger



effects on achievement for TGT and STAD than for the other techniques, particularly when low level skills are assessed. This is probably due to the use in these techniques of a highly structured system of instruction, team tasks, and team rewards. Students in TGT and STAD are given worksheets to study in their teams. They also receive answers to the worksheet items, so that each team has all the resources it needs to make sure that every team member knows the academic material. Also, each student is individually responsible for his or her learning; teammates cannot help each other during the games (TGT) or the quizzes (STAD). In contrast, studies in the Johnsons' classrooms handed in a single worksheet each day that represented the team product, and the steps that lead to that product were left up to the students, so that this technique is low in individual responsibility for learning and low in structure. SmallGroup Teaching is somewhat more structured, and, by encouraging students to establish a division of labor, still has an element of individual accountability for learning (but not nearly as much as in TGT and STAD). Jigsaw is very tightly structured in terms of learning activities, but the original Jigsaw model forces students to rely on one another totally to learn each other's sections. If a student does a poor job of presenting his or her section to the group, the section just does not get learned. The difference between TGT, STAD, and the Hamblin techniques on the one hand, and the other techniques on the other, is a conceptually interesting one. TGT, STAD, and the Hamblin techniques explicitly use the team structure as a motivational device—a reward system that motivates students in their teams to go over academic material again and again until they and their teammates know it. They do not use the team structure as a way to pool individual skills and ideas. The other techniques use the team structure primarily as a. facilitative device, to encourage students to share ideas, to brainstorm, to decide how to structure its own activity, and so forth. TGT, STAD, and the Hamblin techniques have much in common with group contingencies used in behavior modification, while the other techniques, particularly the Johnson techniques and Small-Group Teaching, have more in common with the open school or other humanistic educational programs. Much of the successful laboratory research on cooperation, including that by Deutsch (1949), has involved high level skills, such as problem solving and brainstorming. For these tasks, group membership clearly has a facilitative effect, as the number and quality of ideas produced may be increased by group interaction. For learning how to multiply fractions or punctuate sentences, however, the group might have only a minimal facilitative effect; the important issue is how to get individuals to practice these skills until they master them. The group that is organized as a discussion group might be a poor structure for learning basic mathematics, as Johnson, Johnson, and Scott (1978) discovered. On the other hand, it may be an excellent structure for learning high level cognitive skills, as found by Sharan, Hertz-Lazarowitz, and Ackerman (in press). High level skills have not been specifically measured in the TGT and STAD studies, but it is unlikely that such effects would be found; these techniques focus heavily on the basic skills, such as computation, punctuation, and vocabulary. Another factor that may be systematically related to study outcomes, even within techniques, is the population. In general, studies in middle class white schools have been somewhat less successful in showing achievement gains than those in integrated and/or working class schools. This is further borne out by the findings of Lucker et al. (1976), Slavin (Note 17), and Slavin and Oickle (Note 11) of an interaction between ethnicity and treatment favoring gains in cooperative classes by minority

335


ROBERT E. SLAVIN

students. Wheeler's (Note 1) finding of a strong interaction between cooperative predisposition and achievement gains emphasizes the importance of the particular population involved in a study, particularly as many studies have shown that white, middle class Americans are less cooperatively predisposed than other groups, including American blacks and Mexican-Americans (Madsen & Shapira, 1970). Only one of the 25 studies reviewed here showed greater achievement in a control group than in a cooperatively taught group, and several studies in white middle class settings showed positive achievement effects, so it is unlikely that these techniques would hurt the achievement of white middle class students. If the ethnicity x treatment interaction holds up, cooperative classroom techniques may be a means of reducing the achievement gap between white and minority students while still increasing the achievement of the whites more than in traditional classrooms. However, the contrary evidence is not inconsiderable; study No. 14, one of the largest and best-controlled of the ST AD studies, took place in working class, integrated (40 percent black) junior high schools, and no achievement effects or race x treatment interactions were found. The use of group competition appears to have some positive effect on achievement, although it is confounded with other factors. What may be important is less whether the groups are in competition or not, but whether or not there is an explicit group reward. The Hamblin techniques did not use competition but did use tangible rewards based on the group score, and found positive effects on achievement, while the Johnson techniques did not use any concrete group rewards and did not find positive effects. On the other hand, Small-Group Teaching did not use group competition or explicit group rewards, but did find positive effects on achievement. One additional factor that differentiates the various techniques is the use or nonuse of training of teachers and students in group processing skills. The Jigsaw, SmallGroup Teaching, and Johnson studies in particular place a heavy emphasis on this training, while TGT, STAD, and the Hamblin techniques do no group process training at all. From the pattern of results in these studies, it does not appear that group process training is a useful addition to a cooperative learning model, but this needs further study. Race relations. The effects of student teams on interracial friendship and related variables may be the most important of the outcomes of cooperative techniques. They are relatively consistent and often quite striking. Further, while there may be many ways of increasing student achievement other than the use of student teams, there is some evidence that fostering interracial cooperation is by far the most effective means of improving racial attitudes and behaviors in schools. Slavin and Madden (1979) conducted a secondary analysis of data collected in a national sample of high schools by the Educational Testing Service (Forehand, Ragosta, & Rock, Note 19), and found that teacher workshops, multiethnic texts, minority history, heterogeneous groups, and classroom discussions of race relations had very few effects on students' racial attitudes and behaviors. On the other hand, assignment of students of different races to work with each other and participation of students on multiracial sports teams had strong and consistent effects on race relations. There is not yet enough evidence to differentiate among the different techniques or studies for effects on race relations. Only two studies have failed to find positive effects. Weiger, Wiser, and Cook (1975) found increased positive evaluations of Mexican-Americans by whites in a study using a general cooperative technique, but did not find changes in perceptions of or by blacks or in perceptions of whites by 336



Mexican-Americans. The authors ascribe the failure to find these effects to possible ceiling effects, but it is also possible that their intervention was not structured enough to make the group system salient to the students. One of the four TGT studies (No. 2) failed to find an effect on cross-racial friendship, even though significant effects were found on cross-racial helping. These effects did not appear to be due to a ceiling effect, and are surprising given that three of the successful studies (Nos. 1, 3, and 12) took place in the same school (but with different students and teachers). Weigel et al. also suggest that the use of group competition may have a deleterious effect on race relations, but the positive findings in the TGT and STAD studies do not support this theory. While none of the cooperative learning techniques are demonstrably more effective than others in improving race relations, the TGT and STAD effects are by far the best established. Of the 11 studies that measured race relations, seven evaluated TGT and STAD; of the four remaining studies, one (No. 18) did not use an adequate measure of race relations, and two (Nos. 19 and 27) found effects for white attitudes toward Mexican-Americans only. On the other hand, positive effects were found in six of the seven TGT-STAD studies. The general applicability of the findings is enhanced by the wide variations in types of schools and percent minority in the schools, as well as the repeated positive findings. Mutual concern. As with race relations, the effects of cooperative learning techniques on mutual concern have been generally quite positive, and there are no obvious methodological or technique differences in the effects. The two TGT studies that fail to find effects on mutual concern (Nos. 6 and 7) both took place in thirdgrade classrooms, where students were initially high on the measure, suggesting a possible ceiling effect, but this was not true in studies Nos. 12 and 15, and 16, which also found no effects on this variable. Implications Presently, the research on cooperative learning in classrooms justifies the following conclusions: 1. For academic achievement, cooperative learning techniques are no worse than traditional techniques, and in most cases they are significantly better. 2. For low level learning outcomes, such as knowledge, calculation, and application of principles, cooperative learning techniques appear to be more effective than traditional techniques to the degree that they use: (a) A structured, focused, schedule of instruction; (b) Individual accountability for performance among team members; (c) A well-defined group reward system, including rewards or recognition for successful groups. 3. For high level cognitive learning outcomes, such as identifying concepts, analysis of problems, judgment, and evaluation, less structured cooperative techniques that involve high student autonomy and participation in decision-making may be more effective than traditional individualistic techniques. 4. Cooperative learning techniques have strong and consistent effects on relationships between black, white, and Mexican-American students. 5. Cooperative learning techniques have fairly consistent positive effects on mutual concern among students regardless of the specific structure used. 337


ROBERT E. SLAVIN

6. There is some indication that cooperative learning techniques can improve students' self-esteem. 7. Students in classes using cooperative learning generally report greater liking of school than do traditionally taught students. What these results indicate is that cooperative learning techniques can achieve both cognitive and affective goals, but that there is still much to be discovered about when they do so—for which kinds of students, under what conditions, in which subjects, and for which techniques or components of techniques are positive effects likely to be observed? There is already enough evidence from field research to support the use of cooperative learning techniques in schools, particularly desegregated schools. Many of the techniques are not difficult to learn—for example, teacher training for the STAD studies has taken no more than 3 hours, and often less. All of the techniques can be used by individual teachers in traditionally structured classes without outside help, additional expense, or radical changes in schedules or other external features of school organization. The combined program, as well as work by Johnson and by Sharan outside of research studies, have indicated that cooperative learning techniques can be used in classrooms as the dominant instructional mode. That is, cooperation need not be a supplement to the traditional competitive classroom, but can be used to largely supplant it (although both the combined program and Johnson's programs still contain individualistic and competitive elements). Next Steps The next steps in research on cooperative learning should be directed at explicating the conditions under which these techniques can maximally influence student outcomes. The comparison of techniques presented in this paper is no substitute for a program of research in which important factors are systematically varied in the same design. Some of the factors that might lead to improved understanding of cooperative learning are listed below: 1. What are the effects of group competition as opposed to non-competition, especially on race relations and mutual concern? 2. What are the effects of teacher and student training in group processing skills? 3. How important are explicit group rewards in improving performance and cohesiveness outcomes? 4. How important are racial and ethnic differences in determining the effectiveness of cooperative learning techniques? 5. For which kinds of learning are cooperative techniques most likely to represent an improvement over traditional techniques, and which techniques are most appropriate for which kinds of learning? 6. How important is individual accountability in increasing the effects of cooperative learning on achievement? 7. What are the effects of high student autonomy on outcomes of cooperative learning? 8. What are the long-term effects of cooperative learning on achievement and positive race relations? 9. What are the important components of cooperative learning for increasing students' self-esteem? 338


COOPERATIVE

LEARNING

10. What are the effects of cooperative learning on teachers' role perceptions, attitudes toward children, and other attitudes? 11. How can cooperative learning be adapted for such specialized uses as: (a) mainstreaming, (b) individualization of instruction, (c) bilingual education, (d) remedial education. Of course, there are many other issues, both theoretical and applied, that should be addressed. Cooperative learning represents a substantial change in the technology of classroom instruction. It usually involves simultaneous changes in the reward, task, and authority structures of the classroom. Explicating the consequences of these changes, as well as interactions of the changes with participant and setting characteristics, will be an enormous job, but the results that have been obtained to date indicate that it is a job worth doing. Research on cooperative learning techniques represents an unusual event in the history of educational research. The techniques arose out of social psychological theory; they have been evaluated in numerous field experiments that were generally high in both internal and external validity; and they are in use in hundreds of classrooms across the country and in Israel. As in any program of research there is a need for further investigations of interactions, limitations, and extensions of findings, but the basic model has been validated in classroom settings. Reference Notes 1. Wheeler, R. Predisposition toward cooperation and competition: Cooperative and competitive classroom effects. Paper presented at the Annual Convention of the American Psychological Association, San Francisco, 1977. 2. Hansell, S., & Slavin, R. Cooperative learning and interracial friendships. Paper presented at the Annual Convention of the American Psychological Association, New York, 1979. 3. Gonzales, A. Classroom cooperation and ethnic balance. Paper presented at the Annual Convention of the American Psychological Association, New York, 1979. 4. Edwards, K. J., & Devries, D. L. The effects of Teams-Games-Tournament and two structural variations on classroom process, student attitudes, and student achievement (Report No. 172). Center for Social Organization of Schools, The Johns Hopkins University, 1974. 5. Hulten, B. H., & DeVries, D. L. Team competition and group practice: Effects on student achievement and attitudes (Report No. 212). Center for Social Organization of Schools, The Johns Hopkins University, 1976. 6. DeVries, D. L., Edwards, K. J., & Wells, E. H. Team competition effects on classroom group process (Report No. 174). Center for Social Organization of Schools, The Johns Hopkins University, 1974. 7. DeVries, D. L., & Mescon, I. T. Teams-Games-Tournament: An effective task and reward structure in the elementary grades (Report No. 189). Center for Social Organisation of Schools, The Johns Hopkins University, 1975. 8. DeVries, D. L., Mescon, I. T., & Shackman, S. L. Teams-Games-Tournament in the elementary classroom: A replication (Report No. 190). Center for Social Organization of Schools, The Johns Hopkins University, 1975. 9. DeVries, D. L., Mescon, I. T., & Shackman, S. L. Teams-Games-Tournament (TGT) effects on reading skills in the elementary grades (Report No. 200). Center for Social Organization of Schools, The Johns Hopkins University, 1975. 10. DeVries, D. L., Lucasse, P. R., & Shackman, S. L. Small group versus individualized

339


R O B E R T E. SLAVIN

11.

12.

13.

14.

15.

16.

17.

18.

19.

instruction: Afield test of their relative effectiveness. Paper presented at the annual convention of the American Psychological Association, New York, 1979. Slavin, R. E., & Oickle, E. Effects of learning teams on student achievement and race relations in a desegregated middle school. Paper presented at the annual meeting of the American Educational Research Association, Boston, 1980. Madden, N., & Slavin, R. Cooperative learning and social acceptance of mainstreamed academically handicapped students. Center for Social Organization of Schools, The Johns Hopkins University, 1980. Slavin, R. E., & Karweit, N. An extended cooperative learning experience in elementary school. Paper presented at the annual convention of the American Psychological Association, New York, 1979. Geffner, R. The effects of interdependent learning on self-esteem, interethnic relations, and intra-ethnic attitudes of elementary school children: A field experiment. Unpublished doctoral dissertation, University of California, Santa Cruz, 1978. Bridgeman, D. The influence of cooperative, interdependent learning on role taking and moral reasoning: A theoretical and empirical field study with fifth grade students. Unpublished doctoral dissertation, University of California, Santa Cruz, 1977. Hertz-Lazarowitz, R., Sharan, S., & Steinberg, R. Classroom learning style and cooperative behavior of elementary school children. Unpublished manuscript, Haifa University (Israel), 1978. Armstrong, B., Balow, B., & Johnson, D. W. Cooperative goal structure as a means of integrating learning disabled with normal progress elementary school pupils. Unpublished manuscript, University of Minnesota, 1977. Slavin, R. E. Student learning team techniques: Narrowing the achievement gap between the races (Report No. 228). Center for Social Organization of Schools, The Johns Hopkins University, 1977. Forehand, G., Ragosta, M., & Rock, D. Conditions and processes of effective school desegregation (Final Report). U.S. Office of Education, Contract No. OEC-0-73-6341. Princeton, New Jersey: Educational Testing Service, 1976.

References ARONSON, E. The Jigsaw classroom. Beverly Hills, California: Sage Publications, 1978. ATKINSON, J. W. Towards experimental analysis of human motivation in terms of motives, expectancies, and incentives. In J. W. Atkinson (Ed.), Motives in fantasy, action and society. Princeton, N.J.: Van Nostrand, 1958. BLANEY, N. T., STEPHAN, S., ROSENFIELD, D., ARONSON, E., & SIKES, J. Interdependence in the

classroom: A field study. Journal of Educational Psychology, 1977, 69(2), 121-128. COLEMAN, J. S. The adolescent society. New York: The Free Press of Glencoe, 1961. COOPER, L., JOHNSON, D. W., JOHNSON, R., & WILDERSON, F. Effects of cooperative, compet-

itive, and individualistic experiences on interpersonal attraction among heterogeneous peers. Journal of Social Psychology, in press. COOPERSMITH, S. Coopersmith self-esteem inventory. San Francisco: Self-Esteem Institute, 1975. DEUTSCH, M. An experimental study of the effects of cooperation and competition upon group process. Human Relations, 1949, 2, 199-231. DEVRIES, D. L., & EDWARDS, K. J. Learning games and student teams: Their effects on classroom process. American Educational Research Journal, 1973, 10, 307-318. DEVRIES, D. L., EDWARDS, K. J., & SLAVIN, R. E. Biracial learning teams and race relations in the classroom: Four field experiments using Teams-Games-Toumament. Journal of Educational Psychology, 1978, 70, 356-362. DEVRIES, D. L., & SLAVIN, R. E. Teams-Games-Toumament: A research review. Journal of Research and Development in Education, 1978, 12, 28-38.



EDWARDS, K. J., DEVRIES, D. L., & SNYDER, J. P. Games and teams: A winning combination. Simulation and Games, 1972, 3, 247-269. GERARD, H. B., & MILLER, N. School desegregation: A long-range study. New York: Plenum Press, 1975. GRAUBARD, P. S. Utilizing the group in teaching disturbed delinquents to learn. Exceptional Children, 1969, 36, 267-272. GROSSACK, M. M. Some effects of cooperation and competition upon small group behavior. Journal of Abnormal & Social Psychology, 1954, 49, 341-348. HAMBLIN, R. L., HATHAWAY, C , & WODARSKI, J. S. Group contingencies, peer tutoring, and accelerating academic achievement. In E. Ramp & W. Hopkins (Ed.), A new direction for education: Behavior analysis. Lawrence, Kansas: The University of Kansas, Department of Human Development, 1971. HAMMOND, L. K., & GOLDMAN, M. Competition and non-competition and its relationship to individual and group productivity. Sociometry, 1961, 24, 46-60. JOHNSON, D. W., & JOHNSON, R. T. Instructional goal structure: Cooperative, competitive, or individualistic. Review of Educational Research, 1974, 44, 213-240. JOHNSON, D. W., & JOHNSON, R. T. Learning together and alone. Englewood Cliffs, N.J.: Prentice-Hall, 1975. JOHNSON, D. W., JOHNSON, R., JOHNSON, J., & ANDERSON, D. The effects of cooperative vs.

individualized instruction on student prosocial behavior, attitudes toward learning, and achievement. Journal of Educational Psychology, 1976, 68, 446-452. JOHNSON, D. W., JOHNSON, R., & SCOTT, L. The effects of cooperative and individualized

instruction on student attitudes and achievement. Journal of Social Psychology, 1978, 104, 207-216. JOHNSON, D. W., & JOHNSON, S. The effects of attitude similarity, expectation of goal facilitation, and actual goal facilitation on interpersonal attraction. Journal of Experimental Social Psychology, 1972, 8, 197-206. JONES, S. C , & VROOM, V. H. Division of labor and performance under cooperative and competitive conditions. Journal of Abnormal and Social Psychology, 1964, 68(3), 313-320. JULIAN, J. W., & PERRY, F. A. Cooperation contrasted with intra-group and inter-group competition. Sociometry, 1967, 30, 79-90. LOTT, A. J., & LOTT, B. E. Group cohesiveness as interpersonal attraction: A review of relationships with antecedent and consequent variables. Psychological Bulletin, 1965, 64, 259309. LUCKER, G. W., ROSENFIELD, D., SIKES, J., & ARONSON, E. Performance in the interdependent

classroom: A field study. American Educational Research Journal, 1976, 13, 115-123. MADSEN, M. C , & SHAPIRA, A. Cooperative and competitive behavior of urban Afro-American, Anglo-American, Mexican-American, and Mexican village children. Developmental Psychology, 1970, 3, 16-20. MALLER, J. B. Cooperation and competition. New York: Teachers College, Columbia University, 1929. MICHAELS, J. W. Classroom reward structures and academic performance. Review of Educational Research, 1977, 47(1), 87-98. SHARAN, S., HERTZ-LAZAROWITZ, R., & ACKERMAN, Z. Learning in cooperative small groups and academic achievement of elementary school children. Journal of Experimental Education, in press. SHARAN, S., & SHARAN, Y. Small-Group Teaching. Englewood Cliffs, N. J.: Educational Technology Publications, 1976. SHERIF, M., & SHERIF, C. Groups in harmony and tension. New York: Harper, 1953. SLAVIN, R. E. Classroom reward structure: An analytic and practical review. Review of Educational Research, 1977, 47(4), 633-650. (a) SLAVIN, R. E. HOW student learning teams can integrate the desegregated classroom. Integrated Education, 1977, 75(6), 56-58. (b)


R O B E R T E.

SLAVIN

SLAVIN, R. E. A student team approach to teaching adolescents with special emotional and behavioral needs. Psychology in the Schools, 1977, I4(\), 77-84. (c) SLAVIN, R. E. Student teams and comparison among equals: Effects on academic performance and student attitudes. Journal of Educational Psychology, 1978, 70, 532-538. (a) SLAVIN, R. E. Student teams and achievement divisions. Journal of Research and Development in Education, 1978, 12, 39-49. (b) SLAVIN, R. E. Using student team learning. Baltimore, MD: Center for Social Organization of Schools, The Johns Hopkins University, 1978. (c) SLAVIN, R. E. Effects of biracial learning teams on cross-racial friendships. Journal of Educational Psychology, 1979, 71, 381-387. SLAVIN, R. E. Effects of student teams and peer tutoring on academic achievement and time on-task. Journal of Experimental Education, in press. SLAVIN, R. E., & MADDEN, N. A. School practices that improve race relations. American Educational Research Journal, 1979, 16, 169-180. THOMAS, E. J. Effects of facilitative role interdependence on the group functioning. Human Relations, 1957, 10, 347-366. WALBERG, H. J., & ANDERSON, G. J. Classroom climate and individual learning. Journal of Educational Psychology, 1968, 59, 414-419. WEIGEL, R. H., WISER, P. L., & COOK, S. W. Impact of cooperative learning experiences on cross-ethnic relations and attitudes. Journal of Social Issues, 1975, i / ( l ) , 219-245. WHEELER, R., & RYAN, F. L. Effects of cooperative and competitive classroom environments on the attitudes and achievement of elementary school students engaged in social studies inquiry activities. Journal of Educational Psychology, 1973, 65, 402-407.

AUTHOR R O B E R T E. SLAVIN, Research Scientist, Center for Social Organization of Schools, Johns Hopkins University, Baltimore, Md. 21218. Specialization: Social psychology of education; motivation; field research methodologies.
