Professional Documents
Culture Documents
ETR&D,Vol.39,No,3,pp. 27-40ISSN1042-1629 27
28 ~R&D, Vol. 39, No. 3
of 0.66 favoring cooperative methods (John- another. For example, Beane and Lemke (1971)
son & Johnson, 1989). An analysis of more found that heterogeneous ability grouping
than 226 studies comparing the achievement improved the achievement of the most able
resulting from cooperative versus individual group members at the expense of the least
learning produced an effect size of 0.63 in favor able. However, Webb (1982) reported that het-
of cooperative methods. erogeneous ability grouping was effective for
Cooperative learning also has important both high- and low-ability students. When
social benefits (Slavin & Oickle, 1981). Accord- grouped homogeneously, high-ability stu-
ing to Sharan and associates (Sharan, 1980; dents may mistakenly assume that everyone
Sharan, KusseU, Sharan, & Bejarano, 1984), understands and therefore may interact less
students who work cooperatively improve effectively than when grouped heterogene-
both their self-esteem and attitudes toward ously. Low-ability students may simply be
peers and school work, are more altruistic incapable of supporting each others' learning
toward group members than to other students needs (Webb, 1988). Another perspective, one
in class, and display improved relations with commonly expressed among teachers, sug-
group members of diverse races and cultures. gests that the most able students learn more
Typically, cooperative learning advocates slowly, cover less information, and are not
recommend that students be grouped heter- challenged academically when grouped with
ogeneously, that is, that group composition less able peers, resulting in reductions in both
be manipulated to include students with efficiency and magnitude of learning.
diverse experiences. The rationale for heter- Recently, the effects of group ability composi-
ogeneous grouping is based on both affective tion on learning have been further investigated
and cognitive considerations. Students en- by Hooper and Hannafin (1988). High- and
counter wider diversity in heterogeneous than low-ability eighth-grade students, grouped
in homogeneous groups. Thus, heteroge- heterogeneously or homogeneously by ability,
neous grouping is more likely to improve inter- completed a mathematics tutorial and an-
personal attraction among group members. swered a series of posttest items. Results indi-
Heterogeneous grouping can help to disman- cated that heterogeneous grouping increased
fie social barriers and misconceptions between the achievement of low-ability students by
majority and minority groups, the handi- approximately 50% compared to their homo-
capped and non-handicapped, males and geneously grouped peers. In contrast, homo-
females, and the socially advantaged and dis- geneous grouping increased the achievement
advantaged (Johnson & Johnson, 1989). of the high-ability students by approxi-
Advocates claim that heterogeneous ability mately 12% compared to their heterogene-
grouping benefits both high- and low-ability ously grouped counterparts. The evidence
students. Less able, or disadvantaged, stu- suggests significant payoffs for low-ability stu-
dents receive more instructional support and dents from heterogeneous groupings, but
regulation from their partners than from the potential decrements in the performance of
classroom teacher, are more actively involved, their high-ability cooperative learning partners.
and may observe their partners' learning strat- Although the effectiveness of cooperative
egies (Swing & Peterson, 1982). Concurrently, learning is often attributed to interaction
more able students may also benefit cogni- among group members (see summary by
tively from explaining lesson concepts to their Webb, 1989), little is known about the rela-
partners and from the opportunity to practice tionship between intra-group interaction and
important skills. achievement. Presumably, achievement dif-
Yet, despite the potential social benefits, the ferences attributable to group composition
cognitive effects of heterogeneous ability correspond to differences in intra-group inter-
grouping have not been established. Some action. Indeed, the nature of intra-group coop-
research indicates that heterogeneous group- eration is potentially of greater importance
ing of high- and low-ability students supports than group composition per se (Webb &
the needs of one group at the expense of Lewis, 1988). It is important, therefore, to
COOPERAI]VECBI 29
determine how group composition influences instructional efficiency would be greater for
intra-group interaction and to develop meth- high-ability students in homogeneous groups
ods that promote successful interaction. than for those in heterogeneous groups.
Ineffective interaction may also influence
affective responses. Allport (1954) noted that
racial prejudice is reduced not simply by social
M~HOD
contact, but by interaction that promotes
knowledge and acquaintance of group mem-
Subjects
bers. Although Allport specifically addressed
the goal of improved race relations, higher lev-
A sample of 125 sixth-and seventh-grade stu-
els of interaction may improve social relations
dents selected from a rural, predominantly
between group members in general. When
white, middle school participated in the study.
students do not interact, or when the amount
Only students with uniformly high or low per-
of cooperation is low, improved attitudes
formance on both the mathematics subtest of
among partners may be hampered.
a standardized achievement test and a study
One method to promote interaction involves
pretest were included. High-ability students
increasing individual accountability, wherein
scored at or above the 70th percentile for math-
each group member must demonstrate mas-
ematics on the California Achievement Test
tery of content embedded in the instruction.
and above the mean on a pretest specifically
Contrasted with deriving a "team response,"
designed for the study; low-ability students
where less able students might simply defer
scored at or below the 40th percentile and
to those who are more able, or more able stu-
below the mean on the pretest. Special edu-
dents may attempt to dominate, individual
cation students were not included in the
accountability may promote qualitatively and
study. Participation was voluntary and sub-
quantitatively superior interaction. Such tech-
ject to parental consent.
niques may be used both to isolate and reme-
diate potential learning problems within the
group and to provide an additional incentive
Materials
to cooperate (Hooper, Ward, Hannafin, &
Clark, 1989). This technique may reduce the
Pretest
damaging "free rider" and "sucker" effects
(Kerr, 1983; Kerr & Bruun, 1983) by motiva- The pretest comprised 30 items, including
ting more able group members to provide both computation problems requiring appli-
help, and less able members to invest suffi- cation of mathematical order of computational
cient mental effort to master instruction. precedence [e.g., (11 - 7) x 2] and problems
The purposes of this study were to examine: requiring both simple algebraic substitution
(1) the effects of heterogeneous and homoge- and computation precedence [e.g., (3a - c) x b].
neous group ability composition on achieve- Students provided short-answer responses to
ment and interaction; (2) the influence of each pretest item. The pretest essentially par-
group composition on instructional efficiency; alleled the symbolic operations required dur-
and (3) the effects of strategies requiring dif- ing the study. The K-R 20 reliability for the
ferent levels of performance accountability. pretest was. 92.
The achievement, interaction, and learning
efficiency of high- and low-ability students
Cooperation Training
were compared in heterogeneous and homo-
geneous groups featuring either individual Training was designed to facilitate effective
or group performance accountability. It was intra-group interaction and cooperation. Both
predicted that low-ability students would domain-general and domain-specific tech-
demonstrate higher achievement and would niques were employed. Domain-general strat-
interact more in heterogeneous than in homo- egies improve the affective atmosphere within
geneous groups. It was also predicted that each group. These techniques generally im-
30 ETR&D,Vol.39, No, 3
instruction. Alternatively, groups were told to their corresponding values and functions are
discuss their errors and to attempt a parallel contained in Figure 1.
version of the quiz when ready. This proce- The second segment emphasized methods
dure continued either until mastery was dem- for applying the symbols when used in com-
onstrated or until the mastery quiz had been binations. Instruction was provided on how
repeated twice. After two repetitions, the next to evaluate strings of symbols, and students
segment was presented. were given examples illustrating the evalua-
The fact segment of the lesson emphasized tion process. Practice was then provided in
learning the basic symbol set. Symbols were the form of three 3-symbol strings, three
presented and students were instructed to 5-symbot strings, and three 7-symbol strings
describe to their partners the relationships (see Figure 2). An embedded mastery quiz
among the symbols. Each symbol represented was presented at the end of this section,
a constant or an operation. The value of a con- which again required 100% accuracy in order
stant was determined by the number of lines to proceed.
comprising the symbol. For example, a square The final segment required that students
represented the value four. Parabolas repre- learn the role of a special modifier (see Fig-
sented operations. For example, the opera- ure 3). The modifier caused symbols con-
tions "add" and "subtract" were represented tained within a box to be evaluated before
by parabolas similar to the "greater-than" and other symbols and then doubled. A series of
"less-than" symbols. However, students were related practice questions was presented, and
not instructed in the rules governing the sym- mastery of five quiz items was needed for
bols. Eight multiple-choice practice items were completion.
then provided. Upon completion, a paper- Two versions of the CBI lesson were de-
and-pencil mastery quiz comprising eight signed to reflect different approaches to
questions was presented. Mastery was set at assessing mastery and readiness to continue
100% due to its prerequisite nature for sub- the instruction, one emphasizing individual
sequent phases. The symbols employed and accountability and the other group account-
Constants Ooerations
Represents 0 "~ RepresentsAdd
Represents 1
Represents Multiply
X Represents2
Represents 3
Represents Subtract
/ Represents 4
II X
Answer: 8
Answer: 2
Answer: 12
x >1 m
Answer: t0
ability for attaining mastery. Students in the short-answer items, including 8 fact, 13 appli-
individual accountability group (IR) cooper- cation, 9 problem-solving, and 10 generaliza-
ated during the lesson, but answered the quiz- tion questions. Fact and application questions
zes individually without conferring with their reflected the lesson content. Fact questions
partner. When both students completed a quiz, measured recall of basic symbol meanings,
they exchanged and scored each other's quiz and application questions measured the abil-
booklets. The lower score obtained by either ity to evaluate strings of symbols. General-
group member constituted the group score and ization questions required students to solve
was used to determine group mastery. Stu- questions using unseen symbols that were
dents in the group accountability group (GR) based upon the same rules as the learned
cooperated during the lesson and collaborated symbol system. For example, a hexagon rep-
to complete and score a single quiz. The score resented the numeral 6. Problem-solving
was used to determine group mastery. questions required students to apply the sym-
bols and operation rules to solve unfamiliar
problems. Samples of each question type are
Posttest
shown in Figure 4.
The posttest contained 40 questions divided The questions were presented in a ten-page
among fact, application, generalization, and booklet. Students were allocated 17 1/2 min-
problem solving. The questions were all utes to complete the posttest: 2 minutes to
COOPERATIVECBI 33
Answer: Add
Answer: 12
Answer: X
complete the answers on each of the first five score, representing the frequency of each
pages and 1 1/2 minutes to complete the group member's interaction, was also record-
answers on each of the remaining pages. To ed. Thus, two students from the same group
avoid providing unnecessary cues, students received different cooperation scores if one
were not permitted to review previous pages student interacted more often than his or her
or advance to subsequent pages until the time partner. The cooperation rate for each student--
had elapsed. The overall K-R 20 posttest reli- the cooperation score divided by the number
ability was 0.89. of observations--was used for the cooperation
analysis. The number of quizzes completed
by each group before demonstrating mastery
Dependent Measures was also recorded as an indication of the rel-
ative efficiency of grouping methods.
Three classes of dependent measures were
obtained: achievement, rate of cooperative
interaction, and number of embedded quiz- Design and Data Analysis
zes completed. The achievement measures
were obtained through the posttest subscales The study employed a 2 x 2 x 2 ( x 4) mixed-
and were an indication of lesson-based learn- effects factorial design. The between-subjects
ing by individual students. The cooperation factors included Ability (high, low); Cooper-
34 ETR~D, Vol,39, No. 3
ABILITY
High (n = 60) M 98.33 87.21 73.00 66.11 81.33
SD 6.29 16.41 24.31 23.36 12.22
Low (n = 65) M 95.39 67.46 44.46 29.74 58.81
SD 9.78 22.25 25.62 24.07 15.79
ACCOUNTABILITY
IR (n = 62) M 97.38 74.07 56.45 43.91 67.54
SD 8.81 22.12 29.09 31.45 18.26
GR (n = 63) M 96.23 80.34 59.84 50.44 71.67
SD 7.98 21.79 28.48 28.07 17.84
36 ~n~D, VoW.39, No. 3
TABLE2 [] MeansandStandardDeviationsoftheCooperationScores
Group
Ability Accountability HETEROGENEOUS HOMOGENEOUS
HIGH
IR M 9.63 10.93
SD 4.40 3.45
n (15) (15)
GR M 13.42 12.73
SD 4.00 5.27
n (14) (16)
Total M 11.46 11.86
SD 4.57 4.50
n (29) (31)
LOW
IR M 8.76 6.50
SD 3.42 2.09
n (15) (17)
GR M 12.47 9.48
SD 4.17 4.32
n (13) (20)
Total M 10.48 8.11
SD 4.17 3.75
n (28) (37)
Note:Cooperationscores represent the averagenumber of instancesof cooperationper 5-
minute observation.
COOPERATIVECBi 37
significantly more than low-ability, homoge- accountability groups prior to mastery (4.73
neously grouped students (mean = 8.11), p vs. 4.16). In the second analysis, the effect
< .050. Differences in cooperation rates for low- for Accountability, F(1,57) = 6.54, p < .013,
ability students in heterogeneous and homo- was significant. Individual accountability
geneous groups represent an effect size of 0.58 groups required more attempts than group
of a standard deviation. accountability groups (means 3.32 and 2.50,
respectively; effect size = 0.68 of a standard
deviation).
Mastery Quizzes
Correlation Analysis
Groups completed at least one but no more
than three quizzes after each of three instruc- Overall, a significant correlation was found
tional segments. Thus, each group completed between posttest achievement and interaction,
at least three but no more than nine quizzes. r(125) = .409, p < .001. Correlations between
Mastery quizzes were analyzed twice. First, posttest achievement and interaction were also
the total number of quizzes completed during examined for high- and low-ability students
the entire lesson was examined. However, in heterogeneous and homogeneous groups.
because a few groups failed to complete the Significant relationships were found for high-
entire lesson, a second analysis was conducted ability students in heterogeneous groups, r(29)
to examine differences across only the first = .472, p < .010, and for low-ability students
two segments, which were completed by in heterogeneous groups, r(28) = .558, p <
all groups.
.002. However, correlations were not signifi-
In both analyses, a significant effect was
cant in homogeneous groups for either high-
found for Instructional Group: first analysis,
ability, r(31) = .186, p > .010, or low-ability
F(2,52) = 4.77, p < .013; second analysis,
students, r(37) = -.027, p > .010. In effect,
F(2,57) = 3.38, p < .041. Follow-up contrasts
cooperation frequency was a good predictor of
for the first analysis indicated that signifi-
student achievement only when students were
cantly more mastery quizzes were attempted
grouped heterogeneously. Further analyses
in homogeneous low-ability groups (mean =
focused on testing the differences between
5.18) than in the heterogeneous and homo-
the correlations obtained by high- and low-
geneous high-ability groups (means 4.11 and
ability students in heterogeneous versus
4.07 respectively), p < .050. This finding re-
homogeneous groups. Using Fisher-Z trans-
flects the higher probability of failure in the
formations, significant differences were found
homogeneous low-ability groups. However,
between low-ability students, Z = 2.28, p <
follow-up contrasts for the second analysis
.030, but not between high-ability students,
revealed that, in addition to greater efficiency
Z = 1.18, p > .050, indicating that the corre-
in heterogeneous than in homogeneous low-
lations between cooperation and achievement
ability groups (means 2.86 and 3.37, respective-
were reliably different only for the low-ability
ly; effect size = 0.43 of a standard deviation),
students.
efficiency was greater in homogeneous high-
ability (mean 2.40) than in heterogeneous
groups (effect size = 0.38 of a standard devi- DISCUSSION
ation), p < .050. Apparently, in the first anal-
ysis, this result was absent due to missing Although little is known about the precise
data from the least efficient of the heteroge- nature of the relationship between coopera-
neous groups. tive interaction and learning, most research-
tn the first analysis, the effect for Accountabil- ers suggest that interaction and achievement
ity, F(1,52) = 3.40, p < .071, approached sig- are positively related. Results from this study
nificance. Overall, individual accountability support this assertion. A significant and pos-
groups required more attempts than group itive overall correlation was found between
38 ETR~D,VoW.39, No. 3
Swing & Peterson, 1982). Without training, neity increases (Johnson & Johnson, 1989).
interaction is often ineffective: Students imi- Consequently, the results may not generalize
tate behavior that is familiar but unrelated to to other kinds of heterogeneous groups, i.e.,
group success (McKellar, 1986). However, con- groups containing students of high and aver-
siderable research is needed to identify inter- age ability, average and low ability, or high,
action variables that mediate learning and average, and low ability. Research is needed
techniques that induce students to interact to examine the effects of these other group
appropriately. compositions.
Ability grouping appears to influence in- Future research should also analyze the
structional efficiency. Low-ability homoge- nature of cooperative interaction. In this study,
neous groups attempted significantly more only an aggregate measure of interaction was
quizzes than other groups. Furthermore, the recorded. Although a stronger relationship
second analysis indicated that heterogeneous between interaction and achievement was
groups attempted more quizzes than high- inferred for heterogeneously than for homo-
ability homogeneous groups. These results geneously grouped students, the quality of
suggest that learning was most efficient for interaction was not measured.
homogeneously grouped, high-ability stu- Small-group models offer considerable po-
dents and least efficient for homogeneously tential for cognitive, affective, and fiscal
grouped, low-ability students; apparently the benefits for CBI. However, group composition
low-ability students benefitted from, but has important implications, especially when
slowed the progress of, the high-ability stu- learning gains for one group are achieved at
dents. Again, however, the effect sizes for both the expense of even small achievement losses
findings were relatively small (less than 0.50 for another group. Equity concerns will con-
of a standard deviation) and therefore may be tinue until a model is shown to maximize
of little practical importance. achievement for students of all abilities. The
Three recommendations for future research results suggest that heterogeneous grouping
should be noted. First, group learning was not provides a supportive learning environment
compared to individualized instruction in this for low-ability students, but that the benefit
study. Although some evidence supports the is partially offset by reduced efficiency for the
efficacy of cooperative learning for students most able students. []
of all abilities, direct comparison of individu-
alized and cooperative CBI was not possible
from the present data. Future research should REFERENCES
include control groups to compare the effects
of ability grouping across group and individ- Allport, G. W. (1954). Nature of prejudice. Cambridge,
ual treatments. Second, findings from this MA: Addison-Wesley.
study may not generalize to larger groups. Stu- Bargh, J. A., & Schul, Y. (1980). On the cognitive
benefits of teaching. Journal of Educational Psychol-
dent motivation may be mediated by group
ogy, 72, 593-604.
size. Free riding is more apt to occur when Beane, W. E., & Lemke, E. A. (1971). Group vari-
groups include more than two persons; thus, ables influencing the transfer of conceptual behav-
the issue of social dilemmas may be more ior. Journa! of Educational Psychology, 62, 215-218.
important in larger groups. Researchers should Cohen, E. G. (1986). Designing groupwork: Strategies
also examine whether free riding is more prev- for the heterogeneous classroom. New York: Teach-
ers College Press.
alent during longer treatments. Free-rider Dalton, D. W. (1990). The effects of cooperative
effects may be less prevalent during a short learning strategies on achievement and attitudes
experiment, in which novelty effects are more during interactive video. Journal of Computer-Based
likely to occur, than during a longitudinal Instruction, 17, 8-16.
study. Finally, in this study heterogeneous Dalton, D. W., Hannafin, M. J., & Hooper, S. (1989).
The effects of individual versus cooperative
groups comprised only high- and low-ability computer-assisted instruction on student perfor-
students. Generally, controversy and argu- mance and attitudes. Educational Technology Re-
mentation increase as intra-group heteroge- search and Development, 37(2), 15-24.
40 ~D, Vol. 39, No. 3
Hooper, S., & Hannafin, M. J. (1988). Cooperative (1990). Collaborating with teachers in the inter-
CBh The effects of heterogeneous versus homo- est of student collaboration. International Journal
geneous grouping on the learning of progres- of Educational Research, I3(1), 41-53.
sively complex concepts. Journal of Educational Salomon, G., & Perkins, D. N. (1989). Rocky roads
Computing Research, 4, 413-424. to transfer: Rethinking mechanisms of a neglected
Hooper, S., Ward, T. J., Hannafin, M. J., & Clark, phenomenon. Educational Psychologist, 24(2),
H. T. (1989). The effects of aptitude composition 113-142.
on achievement during small group learning. Jour- Sharan, S. (1980). Cooperative learning in small
nal of Computer-BasedInstruction, 16, 102-109. groups: Recent methods and effects on achieve-
Johnson, D. W., & Johnson, R. T. (1989). Coopera- ment, attitudes, and ethnic relations. Review of
tion and competition: Theory and research. Edina, Educational Research, 50, 241-271.
MN: Interaction Book Company. Sharan, S., Kussell, P., Sharan, Y., & Bejarano, Y.
Johnson, R. T., Johnson, D. W., & Stanne, M. B. (1984). Cooperative learning: Background and
(1985). Effects of cooperative, competitive, and implementation of this study. In S. Sharan, P.
individualistic goal structures on computer- Kussell, R. Hertz-Lazarowitz, Y. Berano, S. Raviv,
assisted instruction. Journal of Educational Psychol- & Y. Sharan (Eds.), Cooperativelearning in the class-
ogy, 77, 668-677. room: Research in desegregated schools (pp. 1-45).
Johnson, R. T., Johnson, D. W., & Stanne, M. B. HiUsdale, NJ: Lawrence Erlbaum.
(1986). Comparison of computer-assisted coop- Slavin, R. E., & Oickle, E. (1981). Effects of cooper-
erative, competitive, and individualisticlearning. ative learning teams on student achievement and
American Educational ResearchJournal, 23, 382-392. race relations: Treatment by race interactions. Soci-
Kerr, N. L. (1983). Motivation losses in small groups: ology of Education, 54, 174-180.
A social dilemma analysis. Journal of Personality Swing, S. R., & Peterson, P. L. (1982). The relation-
and Social Psychology, 45, 819-828. ship of student ability and small-group interac-
Kerr, N. L., & Bruun, S. E. (1983). The dispensabil- tion to student achievement. American Educational
ity of member effort and group motivation losses: ResearchJournal, 19, 259-274.
Free-rider effects. Journal of Personality and Social Webb, N. M. (1982). Peer interaction and learning
Psychology, 44, 78-94. in small cooperative groups. Journal of Educational
King, A. (1989). Verbal interaction and problem- Psychology, 74, 642-655.
solving within computer-assisted cooperative Webb, N. M. (1987). Peer interaction and learning
learning groups. Journal of Educational Computing with computers in small groups. Computers in
Research, 5, 1-15. Human Behavior, 3, 193-209.
LatanG B., Williams, K., & Harkins, S. (1979). Many Webb, N. M. (1988, April). Peer interaction and learn-
hands make light work: The causes and conse- ing in small groups. Paper presented at the annual
quences of social loafing. Journal of Personality and meeting of the American Educational Research
Social Psychology, 37, 822-832. Association, New Orleans, LA.
McDonald, B. A., Larson, C. O., Dansereau, D. E, Webb, N. M. (1989). Peer interaction and learning
& Spurlin, J. E. (1985). Cooperative dyads: Impact in small groups. International Journal of Educational
on text learning and transfer. Contemporary Edu- Research, 13(1), 21-39.
cational Psychology, 10, 369-377. Webb, N. M., & Lewis, S. (1988). The social con-
McKellar, N. A. (1986). Behaviors used in peer tutor- text of learning computer programming. In R. E.
ing. Journal of Experimental Education, 54,163-168. Mayer (Ed.), Teaching and learning computer pro-
Palincsar, A. S., Stevens, D. D., & Gavelek, J. R. gramming. Hitlsdale, NJ: Lawrence Erlbaum.