Volume 7, Number 17
Submitted: July 16, 2002
First Revision: August 13, 2002
Second Revision: August 13, 2002
Accepted: August 13, 2002
Publication date: August 14, 2002
Elaine Chapman
The University of Sydney
ABSTRACT
Research into the effects of cooperative learning on academic performance has produced conflicting results. This study aimed to assess whether these effects varied with the incentive structure under which groups worked and with the level of social cohesiveness between group members. Eightynine 5^{th} and 6^{th} grade students were assigned randomly to one of four conditions in a 2 (incentive) by 2 (cohesiveness) factorial design. Results indicated that students who received rewards based on their individual contributions to an overall group product outperformed those who received rewards based on an overall group product alone. Students in the former condition also made significantly greater prepost increases on a sociometric scale. In contrast, students who worked in groups that were high in social cohesiveness performed marginally worse than those who worked in low cohesive groups. Implications of these results for theory and practice in the area are discussed.[293]

[294]
Cooperative learning strategies are now widely advocated
as a means by which schools can improve students' social integration (e.g.,
Pettigrew, 1998). Despite this, recent surveys suggest that structured small
group methods have not found widespread application in classroom settings
(Autil, et al., 1998). Despite their positive effects on social and personal
outcomes, research into the effects of these methods on academic performance
has produced conflicting results (Slavin, 1996). Reduced effects of cooperative learning have often been ascribed
to motivational losses that occur in the group process. Examples of such losses include "freerider"
effects, in which some members allow other members to do all the work (e.g.,
Kerr & Brunn, 1983), and "sucker" effects, where highachieving
members reduce their efforts to avoid having to do all the work (Kerr, 1983).
Slavin (1996) has argued that in order to have positive effects
on student achievement, cooperative learning should incorporate two key components:
Group rewards and individual accountability. In this view, members of cooperative groups
should receive rewards based on the performance of their groups as a whole. Slavin argued that without this component,
students would not be motivated to interact effectively on their assigned
tasks. Slavin further stipulated,
however, that group rewards would not be effective in motivating all students unless the performance of groups was explicitly
determined by the individual achievements of group members. Slavin posed that without the latter component,
the positive effects of the group reward system on member motivation would
be lost through diffusion of responsibility amongst group members and resulting
"free rider" and "sucker" effects.
These propositions have been supported through a recent metaanalysis
of cooperative learning evaluations. Slavin (1996) cumulated the effects of 99 studies that compared
the achievement effects of cooperative learning and more traditional individualistic
or competitive instructional approaches. When the approaches were classified
by the criteria outlined above (i.e., inclusion or noninclusion of group
reward and individual accountability components), the median effect size for
approaches that used both group rewards and individual accountability was
0.32, as compared with a median effect size of 0.07 for methods that used
group rewards only or individual accountability only. The median effect size for methods that
did not incorporate either of these components was 0.16.
Other researchers in the field (e.g., Cohen, 1994; Kohn,
1991; Schaps & Lewis, 1991), however, have expressed concerns about the
use of incentives based on individual group member performance. For example, Cohen (1986) states that
"if the task is challenging and interesting, and if students are sufficiently
prepared for skills in group process, students will experience the process
of groupwork itself as highly rewarding...[N]ever grade or evaluate students
on their individual contributions to the group product" (pp. 6970).
Although these researchers do not explicitly reject the notion that
providing group rewards for individual performance will have positive effects
on learning in cooperative groups, they argue that the same effects can be
achieved in other, more desirable ways.
Advocates of the social cohesion perspective (e.g., Johnson & Johnson, 1992; Cohen, 1986; Sharan & Sharan, 1976) argue that the extent to which cooperative group members suffer from the motivational losses described above will depend largely on the cohesiveness of the group. For example, Johnson & Johnson (1994) assert that "...the higher the cohesiveness of a group, the more productive it tends to be," where group cohesiveness is based on "members liking each other, desiring to continue to be part of the group, and being satisfied with their group membership" (p.26). Although the concept of group cohesiveness is broad and multifaceted, this view (i.e., that it is primarily a function of the level of interpersonal attraction between group members) has dominated research into the effects of group cohesiveness on achievement and productivity (Hogg & Abrams, 1988). In summary, advocates of the group cohesiveness perspective pose that motivational losses associated with "freerider" and "sucker" effects are less likely to occur in highly cohesive groups, making the use of group rewards based on the individual learning of group members redundant.
[294]

[295]
The results of two early studies provide some support for
the proposition that students achieve more when they work in highly cohesive
groups. In the first of these (Shaw & Shaw,
1962), secondgraders studied spelling in either high cohesive groups (in
which no member rejected any other member, and some positive choices were
represented) or low cohesive groups (where no member chose any other member,
and some rejections were represented). Groups worked together for three daily
sessions, and completed tests on the material covered on the second and third
days of this study period. A
significant positive correlation between group cohesiveness and achievement
was found in the second, but not in the third session tests.
In the second study (Stam, 1973), fifthgraders completed a
peer nomination scale (e.g., asking them to indicate which members of their
class were their “best friends”) and were then assigned to fourmember
cooperative groups either on the basis of mutual nominations (the high cohesiveness
condition) or on a random basis (the low cohesiveness condition).
Each group then completed both a convergent thinking task (i.e., a
series of arithmetic word problems) and a divergent thinking task (i.e., writing
of a group poem). Stam reported that sociometrically chosen groups performed
significantly better on the divergent thinking task, although there were no
significant differences on the convergent thinking task.
Given these results, it is possible that both group cohesiveness and group reward contingencies may act as moderators of learning in cooperative groups. In highly cohesive groups, the motivational losses described above may be less likely to occur, making the positive effects of using group rewards for individual member contributions to an overall product unnecessary. On the other hand, the use of group rewards based on the individual learning of all students in the group may reduce or prevent the same motivational losses, thus making redundant any positive effects of high cohesiveness amongst group members.
In addition to its theoretical interest, this question has important practical
implications for the use of cooperative learning in school settings. If both procedures (i.e., placing students
into highly cohesive groups and using incentives based on individual contributions
to a group product) can be used to achieve the same outcome (i.e., reducing
motivational losses), educators could select the procedure most consistent
with their own ideologies, intervention goals, and practical constraints.
For example, some teachers may find using group reward contingencies
impractical due to the demands made on their time and resources. On the other
hand, when improving social relationships between students is a primary or
concurrent intervention goal, placing students into highly cohesive groups
(e.g., groups in which students indicate a preference for working together)
would clearly be counterproductive.
The goal of the present study was to determine the effects
of social cohesiveness and incentive structures on cooperative learning outcomes.
Three types of outcome measures were used: Student achievement tests,
a perceived group cohesiveness scale, and sociometric peer ratings.
The latter two measures were included to assess any collateral effects
of the experimental procedures on group processes.
[295]

[296]
METHOD
Participants
Eightynine students from three elementarylevel classes participated in this study. The three classes were drawn from a school in a relatively high socioeconomic Australian suburb. Class A comprised 31 fifthgrade students (17 males, 14 females). Class B, a split fifth/sixthgrade class, included 16 fifthgrade students (10 males, 6 females) and 13 sixthgrade students (8 males, 7 females). Class C included 27 sixthgrade students (13 males, 14 females). Median ages at the start of the intervention were 10 years, 2 months for the fifthgrade sample, and 11 years, 3 months for the sixthgrade sample.
Curriculum MaterialsThe academic tasks were drawn from an experimenterdeveloped
workbook of practice items. These items were organised into 96 skill objectives,
which covered four major areas: Fractions and Decimals; Relations and the
Number System; Measurement, Statistics, and Space, and Story Problems. Within
the workbook, each objective presented a worked example for students to follow
so that they were able to work relatively autonomously during lessons.
Students were also given access to answer books during lessons so that
they could check their own work.
Students worked on their own individualized instructional programs during lessons, which were based on their performance on the curriculumbased pretest described below. As each item in this pretest corresponded to one of the objectives in the workbook, students completed all objectives in the workbook that corresponded to items answered incorrectly or not answered on the pretest. After receiving their marked pretests, students recorded the numbers of all such items on a page that appeared at the beginning of each workbook section, and worked from these lists for the remainder of the fourweek experimental period
Dependent MeasuresSociometric Rating
Scale. The sociometric rating scale was used to assess students’
attitudes towards other members of their class. These scores were then used to assign
students to groups within the high and low cohesiveness conditions, as well
as to assess the impact of the different conditions on ratings between group
members. Initially, both a “play with” and a “work with”
dimension (e.g., see Oden & Asher, 1977) were included. However, as a near perfect correlation
was found between ratings on these two scales across all three classes (rs ³ 0.98), mean scores were obtained
for ratings across the two scales and used both in the assignment of students
to cooperative groups, and in the analysis of the results. A more detailed
description of this scale is included in Appendix A. Scores on each subscale
ranged from 1 to 5.
[296]

[297]
Group Cohesiveness
Scale. To determine whether cooperative groups whose members
chose to work together were perceived as more cohesive than those whose members
explicitly chose not to work together, all students completed an eightitem
group cohesiveness scale at posttest. Statements in this scale are presented
in Appendix B. These items were
adapted from a measure developed by Hinkle et al. (1989) to measure intragroup
identification within groups of college students. Items in the scale measure
a number of perceptions commonly identified as central to the notion of group
cohesiveness. These include students'
desire to maintain ingroup status (Johnson & Johnson, 1994), perceptions
of whether cooperating with the group will enhance the likelihood of personal
goal achievement (Sharan & Sharan, 1976), and emotional ties to the group
(Mudrack, 1989a, 1989b). The wording of the items was modified from the original scale
described by Hinkle et al. to be more readily understood
by primaryaged children. Respondents
rated their agreement with each item on a fivepoint scale (ranging from strongly
agree to strongly disagree). Based on an analysis of pretest scores in this
study, the internal consistency of the overall scale was reasonably high (alpha
= 0.86). Scores on this scale ranged from a minimum of 8 to a maximum of 40.
Math Achievement. The primary measure of mathematics achievement was a 76item, experimenterdeveloped test that assessed all skills covered in the fourweek curriculum unit. Two parallel forms were developed. Some of the items required students to select answers from a number of alternatives (i.e., multiple choice format), but most required worked problem solutions. Each of the items corresponded to one of the objectives in the Project Workbook. The test was administered with a 60minute time limit. To obtain an estimate of the parallelforms reliability of the test, 156 students (81 male, 77 female) from grades 4 to 7 in a separate state primary school completed Forms A and B of the test with a oneweek testretest interval. This analysis indicated high parallel forms reliability estimates for grades 5 and 6 (0.89 and 0.90, respectively). Scores on each form of this test ranged from 0 to 76.
A standardized measure
of mathematics achievement was also used to provide an independent measure
of study outcomes. Tests 2A and
2B from the Progressive Achievement Tests in Mathematics (Australian
Council for Educational Research, 1984a) were selected for their relevance
to the Australian mathematics curriculum. Both tests are in multiplechoice
format (47 items in Test 2A and 57 items in Test 2B), and are administered
with a 45minute time limit. In the 1983 standardization (see Australian Council
for Educational Research, 1984b), KR20 reliabilities of 0.91 to 0.94 were
obtained for these tests in a sample of fifth to seventhgrade students.
Raw scores on Test 2A ranged from 0 to 47, while those on Test B ranged from
0 to 57.
Pretesting for all classes was conducted in two sessions on
consecutive days during the week prior to the start of the intervention. The curriculumbased pretest was given
first, followed by the standardized achievement pretest.
The experimental conditions were implemented during students’
daily (onehour) math lessons, over a period of four school weeks. The experimenter
visited all participating classes during each intervention session to ensure
that the procedures were implemented as prescribed. Two days before the start
of the intervention, students in each class received their marked curriculumbased
pretests and a Project Workbook and were briefed on the basic procedures they
would follow during the fourweek intervention period.
This general briefing session was conducted by the experimenter. Each
week, classes focused on a different topic (e.g, Fractions and Decimals). At the end of the week, all students completed a quiz that
was a parallel form of the relevant section in the pre and posttests.
[297]

[298]
Small prizes (e.g., stickers, pencils) were awarded on the
basis of making a threepoint improvement on the score obtained for that section
of the pretest. This improvement system was used to ensure that all students
(regardless of initial ability level) would have an equal opportunity to contribute
to their group’s performance (see Slavin, 1996). Weekly certificates and prizes were awarded
publicly on the day after the endofweek quiz for students in all conditions
(just prior to the beginning of the next experimental session), and were called
in alphabetical order. These rewards were used to establish the two incentive
conditions described in the next section.
Posttesting was conducted over three consecutive days immediately after the last day of the intervention. The Group Cohesiveness and the sociometric scales were completed in the first session, followed by the curriculumbased and the standardized achievement tests, respectively
Experimental ConditionsFollowing the pretesting sessions, students in each of the three classes were assigned randomly (stratifying for curriculum based pretest scores and sex) to either the high or the low group cohesiveness condition. Thus, approximately half of the students in each class were assigned to each condition.
In the high cohesiveness condition, students
were assigned to fourmember cooperative groups so that there were no negative
ratings between members (i.e., no ratings less than 0 on the rating scale),
and each member rated at least one other member of the group positively (i.e.,
a rating of 1 or more on the rating scale).
In the low cohesiveness condition, students
were assigned to groups so that there were no positive choices among members
(i.e., no ratings more than 0 on the rating scale), with each member rating
at least one other member negatively (i.e., a rating of less than 0 on the
pretest rating scale). Assignments
to groups were counterbalanced for sex and scores on the curriculumbased
pretests to ensure that a mixture of males and females and ability levels
was maintained in both high and low cohesive groups.
Within each of these conditions, groups were
then assigned randomly to one of two incentive criteria conditions:
Individual contributions: Students in this condition were encouraged to work together and consult other members in their groups before asking for teacher assistance during lessons. These students were told that they would receive a certificate and a small prize at the end of each week if all members of their group had met their individual performance goals on the endofweek quiz (i.e., made a threepoint improvement on their initial score for the relevant subsection of the pretest). It was made clear that it was the responsibility of all group members to ensure that everyone met their performance goals, and that noone in the group would receive a certificate or prize unless all group members had achieved their respective goals
Group products: Students in this condition followed identical procedures to those in the individual contributions condition during lessons. However, these students did not complete individual quizzes at the end of each week. Instead, each cooperative group completed a parallel form of the curriculumbased pretest as a group on the day before the experiment commenced, and received a group pretest score for the four subtests. During this group pretest, all members were encouraged to contribute to the solution of the test problems, and to reach consensus on each answer before moving onto the next item. Each group also completed an overall group quiz at the end of each week, identical to the endofweek quizzes completed individually by students in the other two conditions. Group members were told that they should aim to contribute to the solution of at least three new items on each endofweek quiz, and that they would receive certificates and prizes if their group as a whole improved by three points on their group pretest score. All group members who were present during the endofweek quiz received a certificate and prize if the group as a whole met its performance goal.
[298]

[299]
Of the 89 students, one was absent during the curriculumbased achievement posttest sessions, while three chose not to complete the sociometric rating scale. As none of these students could be tested at a later date, their data sets were deleted from the relevant analyses. Overall descriptive statistics and bivariate correlations for the dependent measures are presented in Appendix C. For all measures, the experimental conditions were compared using a 2 (incentive criterion: individual contributions versus group products) by 2 (group cohesiveness: high versus low) by 3 (class: A, B, or C) factorial design. As the outcomes were not independent within classes, class was included as an independent variable in all analyses to test for interaction effects with experimental condition. Given the low number of students in each group, multivariate and univariate analyses of variance and covariance were used rather than hierarchical linear models for these analyses. Correlations between individual students’ scores and group mean scores within conditions were nonsignificant for all dependent measures (ps > 0.10), providing some support for conformity to the independence assumption
Group Cohesiveness OutcomesScores on the group cohesiveness scale were subjected to a
univariate analysis of variance (ANOVA). Means and standard deviations for
posttest scores on the Group Cohesiveness Scale are shown in Table 1. The 2 (group cohesiveness) by 2 (incentive
criteria) by 3 (class) ANOVA on these scores indicated no significant main
effect for class (F(2,77) = 1.90, p
> 0.05), for incentive criteria, or for group cohesiveness (both Fs < 1). All twoway interactions were also nonsignificant
(a = 0.05).
Table 1. Means and Standard Deviations for Scores on the Group Cohesiveness Scale By Incentive and Cohesiveness Condition
Incentive Condition 
Cohesiveness Condition 
Group Cohesiveness Scale 

N 
Posttest Mean 

Individual Contributions 
Low 
22 
25.45(6.90) 
High 
23 
25.48(6.08) 

Total 
45 
25.47(6.42) 

Group Products 
Low 
20 
23.15(6.20) 
High 
24 
25.29(6.50) 

Total 
44 
24.32(6.39) 

Collapsed Group Means 
Low 
42 
24.36(6.60) 
High 
47 
25.38(6.23) 

Total 
89 
24.90(6.39) 
[299]

[300]
There was, however, a significant threeway interaction (F(2,77) = 4.84, p < 0.05). Simple
interaction tests performed for each level of class indicated a significant
interaction between performance criteria and group cohesiveness in class B
(F(1,77) = 9.72, p
< 0.016, partial eta squared = 0.31),
but not in either of the other classes (ps > 0.016). Cell
mean contrasts indicated that the effect of group cohesiveness was significant
in the group products, but not in the individual contributions, incentive
condition (F(1,77) = 8.62, p
< 0.008, partial h^{2}
= 0.48; F(1,77) = 2.24, p > 0.008, respectively). This indicates that students in high cohesive groups reported
higher cohesiveness scores than those in the low cohesive groups in the group
products condition in Class B.
The fact that there was no significant main effect for condition
on the group cohesiveness scale does suggest that members of high cohesive
groups did not actually perceive their groups to be more cohesive than those
in low cohesive groups. Thus, while the groups differed initially in the extent
to which members liked one another,
this did not lead to a more general perception of high cohesiveness.
Scores on the two achievement tests (curriculumbased and
standardized) were subjected to a multivariate analysis of covariance (MANCOVA),
with corresponding pretest scores entered as covariates. Screening procedures
for conformity to univariate and multivariate analysis of variance assumptions
produced satisfactory results. Pretest means and standard deviations for the
curriculumbased and standardized tests are presented in Tables 2 and 3, respectively.
A 2 (performance criteria) by 2 (cohesiveness) by 3 (class) multivariate analysis
of variance (MANOVA) on these scores indicated a significant main effect for
class (V = 0.23, F(4,152)
= 4.84, p < 0.05), with univariate analyses of variance (ANOVAs)
indicating significant effects for both the curriculumbased and standardized
tests (F(2,76) = 11.04, p
< 0.025, partial eta squared = 0.22;
F(2,76) = 7.45, p < 0.025, partial eta squared = 0.16, respectively). There
were, however, no significant main effects either for group cohesiveness (V
= 0.01, F(2,75) < 1) or incentive criteria (V =
0.01, F(2,75)
< 1), and all multivariate two and threeway interactions were also nonsignificant
(a = 0.05). Thus,
pretest scores did not differ significantly across the experimental conditions,
either within classes or across the full threeclass sample.
[300]

[301]
Table 2. Means and Standard Deviations for Scores on the CurriculumBased Test by Incentive and Cohesiveness Condition
Incentive Condition 
Cohesiveness Condition 
N 
Pretest Mean 
Posttest Mean 
Adjusted Posttest Mean 
Individual Contributions 
Low 
22 
36.79(14.75) 
49.14(13.14) 
49.57 
High 
23 
37.52(15.34) 
47.04(14.60) 
47.55 

Total 
45 
37.17(14.89) 
48.07(13.79) 
48.56 

Group Product 
Low 
19 
33.70(13.56) 
44.82(12.19) 
46.97 
High 
24 
40.54(15.56) 
47.35(15.77) 
44.64 

Total 
43 
37.43(14.92) 
46.23(14.19) 
45.81 

Collapsed Means 
Low 
41 
35.32(14.11) 
47.13(12.74) 
48.27 
High 
47 
39.06(15.36) 
47.20(15.04) 
46.10 

Total 
88 
37.30(14.82) 
47.17(13.94) 
47.18 
[301]

[302]
Table 3. Means and Standard Deviations for Scores on the Standardized Achievement Test By Incentive and Cohesiveness Condition
Incentive Condition 
Cohesiveness Condition 
N 
Pretest Mean 
Posttest Mean 
Adjusted Posttest Mean 
Individual Contributions 
Low 
22 
27.81(9.69) 
31.77(10.20) 
31.90 
High 
23 
26.82(10.08) 
30.52(11.56) 
31.20 

Total 
45 
27.31(9.79) 
31.13(10.81) 
31.55 

Group Product 
Low 
19 
26.40(8.29) 
31.11(11.13) 
33.06 
High 
24 
30.16(11.05) 
33.25(11.98) 
31.07 

Total 
43 
28.45(9.97) 
32.30(11.53) 
32.06 

Collapsed Means 
Low 
41 
27.14(8.97) 
31.46(10.51) 
32.48 
High 
47 
28.53(10.61) 
31.91(11.73) 
31.14 

Total 
88 
27.87(9.84) 
31.70(11.12) 
31.81 
[302]

[303]
The multivariate test for the relationship between combined pre and posttest scores was significant (V = 0.97, F(4,148) = 34.76, p < 0.05). Univariate regressions indicated significant relationships between combined pretest scores and posttest scores on both the curriculumbased (F(1,74) = 221.90, p < 0.025, partial eta squared = 0.86) and the standardized achievement tests (F(1,74) = 101.77, p < 0.05, partial eta squared = 0.73). Thus, use of these pretest scores as covariates produced a significant reduction in posttest error variance. Tests for heterogeneity of regression across the 12 cells of the design were also nonsignificant (ps > 0.10), indicating that the assumption of regression homogeneity was tenable.
The multivariate test for the threeway interaction was not
significant (V = 0.08, F(4,148) = 1.48, p > 0.05), and there were no significant multivariate
or univariate twoway interactions (all Fs < 1). The
multivariate F for the class
main effect was significant, however (V = 0.36, F(4,148)
= 8.09, p < 0.05), with univariate
ANCOVAs indicating significant effects on both the curriculumbased test (F(2,74) = 18.12, p < 0.025, partial eta squared = 0.33) and the standardized
achievement test (F(2,74) =
8.95, p < 0.025, partial
eta squared = 0.19). Thus, inclusion of class as an independent
variable produced a significant reduction of error variance in achievement
posttest scores.
The MANCOVA indicated a significant main effect for incentive
criterion on adjusted posttest scores (V = 0.12, F(2,73)
= 4.94, p < 0.05). The effect for group cohesiveness also
approached significance (V = 0.06,
F(2,73) = 2.27, p = 0.11). As the adjusted withincells correlation between
curriculum based and standardized posttest scores was significant (r = 0.40, Bartlett’s chisquare(1) = 12.73, p
< 0.05), both univariate and stepdown
Fs were used in the interpretation of these effects.
As the curriculumbased measure provided a direct test of the material
covered in the program, scores on this test were entered prior to standardized
achievement scores in the stepdown analysis. All univariate and stepdown Fs were tested for significance at Bonferroniadjusted
alpha levels to maintain nominal familywise alpha at or below 0.05 for each
set.
Univariate ANCOVAs indicated a significant effect for incentive
criterion on the curriculumbased test (F(1,74) = 7.25, p < 0.025, partial eta squared = 0.09), but not on the
standardized achievement test (F(1,74)
< 1). As indicated in Table
2, the effect on the curriculumbased measure favored the individual contributions
condition. The univariate effect
for group cohesiveness approached significance for scores on the curriculum
based test (F(1,74) = 4.50, p = 0.04, partial eta squared = 0.06) but not for scores
on the standardized achievement test (F(1,74) = 1.33, p > 0.025). As indicated
in Table 2, the marginal effect on the curriculumbased test favored the low
cohesiveness condition.
Means and standard deviations for pre and posttest scores
on the sociometric scale are also shown in Table 4. To determine whether there were significant differences between
the conditions on these scores, a 2 by 2 by 3 ANCOVA was performed.
Again, the ANCOVA indicated a significant relationship between the
pre and posttests (F(1,73) = 63.99, p < 0.0001, partial eta squared = 0.47), but no significant
pretest by condition interaction effects (ps > 0.15), indicating that the use of covariance
analysis was tenable.
[303]

[304]
Incentive Condition 
Cohesiveness Condition 
Sociometric Scale 

N 
Pretest Mean 
Posttest Mean 
Adjusted Posttest Mean 

Individual Contributions 
Low 
21 
2.06(.40) 
2.22(.57) 
2.94 
High 
22 
3.78(.40) 
3.61(.43) 
2.94 

Total 
43 
2.94(.95) 
2.93(.86) 
2.94 

Group Products 
Low 
19 
2.01(.44) 
2.07(.45) 
2.84 
High 
24 
3.72(.39) 
3.38(.48) 
2.76 

Total 
43 
2.94(.95) 
2.80(.80) 
2.80 

Collapsed Group Means 
Low 
40 
2.04(.41) 
2.15(.52) 
2.89 
High 
46 
3.75(.39) 
3.49(.46) 
2.85 

Total 
86 
2.94(.95) 
2.87(.83) 
2.87 
The ANCOVA on sociometric posttests indicated a marginally
significant main effect for incentive condition (F(1,73) = 2.96, p = 0.09 partial eta squared^{ }= 0.04), but all other effects
were nonsignificant (Fs <
1). As indicated by the means
in Table 4, the main effect for incentive indicated significantly higher scores
for the individual contribution condition.
These results indicate that students who worked in cooperative
groups and received group rewards based on individual contributions to the
group product made significantly greater gains on a curriculumbased achievement
test than those who received rewards based an overall group product. Slavin
(1996) predicted that use of group rewards for individual performance would
have positive effects on learning in cooperative groups by eliminating factors
that lead to group motivation losses. As members are made individually accountable for the group’s
success, this precludes any diffusion of responsibility between group members,
eliminating motivational losses associated with "freerider' and "sucker"
effects.
[304]

[305]
The hypothesis that members of groups who chose to work together
would perform better than those in other groups was not supported. Indeed,
it was found that students in high cohesive groups actually performed marginally
worse than those in low cohesive groups.
Thus, the motivational losses discussed above appear to have been more,
rather than less, pronounced in groups where members chose to work together.
Informal observations of high and low cohesive groups suggested that members
of high cohesive groups interacted more than those in low cohesive groups,
but that these interactions were largely offtask. This result is not consistent
with the results reported by Shaw & Shaw (1962) and by Stam (1973). However, as noted, both of these studies
appear to have used group productivity, rather than individual achievement
measures. The discrepancy in
the results of these studies and the present one suggests that although members
who like one another may coordinate their efforts more effectively in a group
task (e.g., a group quiz or assignment), as suggested by the Shaw & Shaw
and Stam results, they may not learn more individually from working with other students that they like.
The results also indicated a significant effect on the sociometric
scale for incentive criterion. Interestingly, this effect favored the individual contributions
condition. It is possible that
this was due to the improvementbased scoring system used in the present study.
When the curriculum materials used are appropriately taskanalyzed
(ensuring that each new skill introduced builds on previous ones), this system
ensures that each member of the group has an equal opportunity to contribute
to the group’s success (Slavin, 1996).
Further, the use of this system imposes a structure that ensures that
the group is not dominated by the contributions of one or two highstatus
members. Gordon Allport (1954) specified that intergroup
contact would promote positive relations between majority and minority group
members only when participants engaged in equal status interaction in the
pursuit of common goals. The result obtained here may simply reflect the fact
that this can only be achieved when group processes are structured by incentives
that impose equal opportunities for all group members to contribute.
The results of this study should, however, be interpreted in
light of the fact that there was no significant difference between the conditions
on the perceived group cohesiveness scale. Thus, while the level of interpersonal attraction between group
members differed initially across the two conditions, this did not lead to
predicted differences in cohesiveness perceptions. This outcome may reflect the nature of the group cohesiveness
scale used. The experimental
manipulation used in the present study focused on social cohesiveness, while the cohesiveness scale focused
more (although not exclusively) on taskbased cohesiveness.
Hackman (1976) defined taskbased cohesiveness as members’ “shared
commitment to the task of the group” (p. 1517). In a metaanalysis of
studies that explored the effects of cohesiveness on group productivity, Mullen
& Cooper (1994) found that, on average, correlational studies revealed
a positive relationship between taskbased cohesiveness and group productivity,
but a negative relationship
productivity and social cohesiveness. Other authors (e.g., Hackman, 1976,
Lott & Lott, 1965) have also posed that increases in interpersonal attraction
will lead to process losses in groups by increasing the number of member interactions
and activities away from the task. Even though these predictions relate specifically to group
productivity outcomes, the same factors may also apply to individual learning
outcomes. As such, further evaluations
could explore the effects of using alternative approaches to increasing cohesiveness
in cooperative groups, such as the task cohesivenessbuilding strategy described
by Johnson et al. (1994), on learning in cooperative groups.
Students in the present study also did not receive any training
in group interaction skills prior to the intervention. Given the emphasis
placed on such training by other researchers (e.g., Cohen, 1994; Johnson et
al., 1994), it is possible that the impact
of group cohesiveness and incentives will vary with students’ preparedness
for cooperative group work. Thus, future evaluations could examine
the effects of these factors on student learning as a function of previous
training in the use of effective cooperative interaction skills. Further,
the present study did not include specific measures of group interaction processes
(e.g., structured observational data). The information yielded by such measures
would greatly facilitate interpretations of overall differences between conditions.
Further work in this area could incorporate the use of schedules similar to
the ones used by Webb (1985), which provide codes for assessing both the quality
and the quantity of interactions between cooperative group members.
[305]

[306]
The results of this study may not generalize across all types
of group learning tasks. In particular,
this study used convergent problemsolving tasks (i.e., problems in which
there was only one correct response), rather than divergent problemsolving
tasks (i.e., problems in which a diversity of solutions are required). It
is possible that the motivational losses described earlier tend to occur more
frequently when groups are assigned to complete convergent tasks.
Because highachieving students are likely to be able to solve convergent
problems more quickly than other members of their groups, lowerachieving
students may be more prone to perceiving their efforts as redundant on these
types of tasks. In contrast, in divergent problem tasks,
a broad range of perspectives is often useful. Thus, future evaluations could assess interactions between
the effects of incentive structure, group cohesiveness, and task structure
on cooperative learning outcomes in the cognitive, affective, and social domains.
Allport, G.W. (1954).
The Nature of Prejudice. Reading, MA: AddisonWesley.
Australian Council for Educational Research. (1984a). The Progressive Achievement Tests in Mathematics. (Forms 1A, 1B, 2A, and 2B.) Hawthorn, Victoria: The Australian Council
for Educational Research.
Australian Council for Educational Research. (1984b). The Progressive Achievement Tests in Mathematics: Teacher’s
Handbook. Hawthorn, Victoria: The Australian
Council for Educational Research.
Autil, L.R., Jenkins, J.R., Wayne, S.K., & Vadasy, P.F.
(1998). Cooperative learning:
Prevalence, conceptualisations, and the relationship between research and
practice. American Educational
Research Journal, 35(3), 419454.
Cohen, E. (1986). Designing
Groupwork: Strategies for the Heterogeneous Classroom. NY:
Teachers College Press.
Cohen, E. (1994). Restructuring
the classroom: Conditions for productive small groups. Review of Educational Research, 64(1), 135.
Hackman, J.R. (1976).
Group influence on individuals.
In M.D. Dunnette (Ed.), Handbook of Industrial and Organisational
Psychology. Chicago: RandMcNally.
Hinkle, S., Taylor, L.A., & FoxCardamone, D.L. (1989).
Intragroup identification and intergroup differentiation: a multicomponent
approach. British Journal of Social Psychology, 28,
305317.
Hogg, M.A., & Abrams, D. (1988). Social Identifications: A Social Psychology
of Intergroup Relations and Group Processes. NY: Routledge.
Johnson, D.W., & Johnson, R.T. (1992). Positive interdependence: Key to effective
cooperation. In R. HertzLazarowitz
& N. Miller, Interaction in Cooperative Groups: The Theoretical Anatomy
of Group Learning. NY:
Cambridge University Press.
[306]

[307]
Johnson, D.W., & Johnson, R.T. (1999). Making cooperative learning work. Theory into Practice, 38(2), 6773.
Johnson, D.W., Johnson, R.T., & JohnsonHolubec, E. (1994).
The New Circles of Learning: Cooperation in the Classroom and School. Alexandria, VA: The Association
for Supervision and Curriculum Development.
Kerr, N.L. (1983). Motivation
losses in small groups: A social dilemma analysis. Journal of Personality and Social Psychology,
45, 819828.
Kerr, N.L., & Brunn, S.E. (1983). Dispensability of member effort and group
motivation losses: Freerider effects. Journal of Personality and Social Psychology, 44,
7894.
Kohn, A. (1991). Group
grade grubbing versus cooperative learning. Educational Leadership, 48(5), 8387.
Lott, A.J., & Lott, B.E. (1976). Group cohesiveness as interpersonal attraction:
A review of relationships with antecedent and consequent variables.
Psychological Bulletin,
64, 259309.
Mudrack, P.E. (1989a).
Defining group cohesiveness: A legacy of confusion? Small Group
Behaviour, 20(1), 3749.
Mudrack, P.E. (1989b).
Group cohesiveness and productivity: A closer look. Human Relations, 9, 771785.
Mullen, B., & Cooper, C. (1994). The relation between group cohesiveness
and performance: An integration. Psychological
Bulletin, 115, 210227.Pettigrew, T.F. (1998). Intergroup contact theory. Annual Review of Psychology,
49, 6585.
Schaps, E., & Lewis, C. (1991). Extrinsic rewards are education’s
past, not its future. Educational
Leadership, 48, 81.
Shaw, M.E., & Shaw, L.M. (1962). Some effects of sociometric grouping upon
learning in a second grade classroom. The Journal of Social Psychology, 57,
453458.
Slavin, R.E. (1992).
When and why does cooperative learning increase achievement? Theoretical and empirical perspectives.
In R. HertzLazarowitz & N. Miller, Interaction in Cooperative
Groups: The Theoretical Anatomy of Group Learning.
NY: Cambridge University Press.
Slavin, R.E. (1996).
Cooperative Learning: Theory,
Research, and Practice. Needham,
MA: Allyn & Bacon.
Sharan, S., & Sharan, Y. (1976). SmallGroup Teaching. NJ: Educational
Technology Publications.
[307]

[308]
Stam, P.J. (1973). The
Effect of Sociometric Grouping on Task Performance in the Elementary Classroom.
Doctoral Dissertation, Stanford University. (University Microfilms Order Number 7330480.)
Webb, N.M. (1985). Student
interaction and learning in small groups. In R.E. Slavin, S. Sharan, S. Kagan, R.
HertzLazarowitz, C. Webb, & R. Schmuck (Eds.), Learning to Cooperate, Cooperating to Learn. NY: Plenum.
A. Sociometric Scale
For the sociometric scale, students were presented with a list
of all other students in their class, and asked to rate extent to which they
would like to work with and to spend their free time with each person named,
for example:


How much would you like to work with this person in the future? 
How much would you like to spend your free time with this person in the future? 

Student Name 
ID 
Not at all 
Not much 
Don’t mind 
A bit 
A lot 
Not at all 
Not much 
Don’t mind 
A bit 
A lot 
John Smith 
## 










[308]

[309]
1. I'm glad I belong to this group.
2. I feel held back by this group.
3. I am an important part of this group.
4. I don't fit in with other kids in this group.
5. I feel strongly tied to this group.
6. I don't think the group is that important.
7. I think this group worked well together.
8. I don't feel
comfortable with the other kids in this group.
C. Means and Bivariate Correlations for Dependent Measures

N 
M 
SD 
1 
2 
3 
4 
5 
6 
7 
1. Group Cohesiveness
Posttest 
89 
6.90 
6.39 
1.00 






2. Sociometric Pretest 
89 
2.95 
0.95 
0.16 
1.00 





3. Sociometric Posttest 
86 
2.87 
0.83 
0.16 
.90 
1.00 




4. CurriculumBased
Achievement Pretest 
89 
37.30 
14.82 
0.20 
0.15 
0.17 
1.00 



5. CurriculumBased
Achievement Posttest 
88 
47.17 
13.94 
.22 
0.04 
0.05 
.90 
1.00 


6. Standardized Achievement
Pretest 
89 
27.88 
9.84 
0.17 
0.08 
0.11 
.81 
.80 
1.00 

7. Standardized Achievement
Posttest 
89 
31.58 
11.11 
0.15 
0.08 
0.07 
.82 
.88 
.82 
1.00 
[309]

[310]
Elaine Chapman (B.A.Hons, Ph.D.) is a Lecturer in the School
of Development
and Learning at the University of Sydney, where she teaches and researches
in research methods, assessment, and educational psychology. Email: e.chapman@edfac.usyd.edu.au
[310]

[311]