Participation and Withdrawal Among Fourth-Grade Pupils

American Educational Research fournal Spring 1992, Vol 29, No. 1, pp. 141-162

Participation and Withdrawal Among Fourth-Grade Pupils

Jeremy D. Finn and Deborah Cox State University of New York at Buffalo

Pupil participation in elementary school classrooms is essential for learning to occur, while nonparticipation in early grades can initiate a cycle that culminates in total withdrawal—dropping out—in later years. This study explores the nature of participation and nonparticipation among fourth-grade students. A sample of 1388 youngsters was rated by their teachers on a questionnaire assessing nonparticipatory behavior, minimal adequate effort, and initiative-taking behavior. Based on the ratings, three groups were formed: active participants, passive participants, and nonparticipating students. The groups were compared on demographic characteristics, attendance, achievement, and self-concept for the the preceding 3 years. Participation groups were clearly distinct on the achievement measures since first grade, and they maintained those distinctions over time. It is concluded that youngsters who withdraw from participation in the classroom should be identified at the earliest possible time to attempt to avert the harmful effects that may ensue.

JEREMY D. FINN is a professor in the Graduate School of Education at the State University of New York (SUNY) at Buffalo, 408 Christopher Baldy Hall, Buffalo, NY 14260. He specializes in multivariate analysis and school and classroom processes.

DEBORAH COX was a graduate research assistant in the Graduate School of Education at SUNY-Buffalo when she worked on this article.

at PENNSYLVANIA STATE UNIV on May 10, 2016http://aerj.aera.netDownloaded from

http://aerj.aera.net

Finn and Cox

T he importance of keeping students actively involved in the education process has been recognized since the time of Dewey:

The active side precedes the passive in the development of the child-nature; . . . the neglect of this principle is the cause of a large part of the waste of time and strength in schoolwork. The child is thrown into a passive, receptive or absorbing attitude. The conditions are such that he is not permitted to follow the law of his nature; the result is friction and waste. (Dewey, 1929, p. 224)

In spite of the obvious importance of students' participation in the classroom, not to mention the larger school environment, the behaviors that may be described as engagement, involvement, or active participation have not been analyzed or studied systematically by educational researchers.

The purpose of this article is to integrate some of the diverse research on students' active participation in school and to describe a study of the forms and antecedents of pupil participation in fourth-grade classrooms. Three research questions are addressed in the study: (1) What specific behaviors distinguish those pupils judged by their teachers to be highly involved in classroom activities from their peers who are more withdrawn? (2) How is participation distributed among children from different races, genders, and socioeconomic groups? And (3) do differences in prior performance distinguish youngsters who are more involved in classroom activities in Grade 4 from their less involved classmates?

Participation in the Elementary Grades

The ways in which students participate in educational endeavors are numerous, even in the early grades. While young children may have little choice about attending school or sitting in an assigned seat, certain degrees of freedom may still be exercised. Simple behaviors such as paying attention and responding to teachers' directions, for example, may be exhibited by some students to a greater extent than others. Research has demonstrated that these behaviors are closely linked to school performance. Attwell, Orpet, and Meyers (1967) found that, of a variety of behaviors observed during a kindergarten ability test, attention was most highly predictive of school achievement in Grade 5. Cobb (1972) observed and coded 14 discrete behaviors exhibited by a large sample of fourth-grade youngsters during arithmetic periods. Attention was the single highest correlate of arithmetic performance, while compliance was high among the remaining significant predictors as well. Lahaderne (1968) observed the behavior of sixth-grade students and tallied their attentive and inattentive behavior. Results indicated that the amount of attentive behavior was positively associated with achievement in reading, arithmetic, and language while inattentive behavior was negatively correlated with achievement.

The extent of students' involvement in classroom activities goes beyond

142



Student Participation and Withdrawal

just the degree of acquiescence, however. McKinney, Mason, Perkerson, and Clifford (1975) observed and coded the behaviors of second-grade youngsters into 12 categories that included attending and passive responding and also more assertive behaviors, such as constructive self-directed activity and aggression. Both of the latter measures were significant correlates of spring performance on the California Achievement Test, while attending and task-oriented interaction approached significance. Swift and Spivack (1969), in an attempt to identify factors that interfere with learning, broadened the scope of observed behaviors still further. After extensive development, their teacher-rated instrument included 45 items that grouped into 13 factors. Several of these factors related to youngsters' performance characteristics (reasoning ability, originality, general anxiety, expressed inability, lack of intellectual independence, and dogmatic-inflexible). The other factors described the form and extent of the youngsters' involvement in classroom activities: verbal interaction, rapport with teacher, disturbance-restlessness, quiet-withdrawn, poor work habits, verbal negativism, and anxious producer. Over 1500 students, ages 12 to 19, in regular classes and in special classes for the emotionally disturbed were rated on the instrument. For all 13 factors, correlations with school grades were statistically significant and substantial; for example, the correlations of the involvement scales with grades for youngsters in regular classes ranged from .26 to .54 in magnitude, all in the expected direction.

Several of the items on the Swift-Spivack questionnaire assessed whether the youngster went beyond the basic requirements of the school and classroom, for example, "raises hand to answer a question or volunteer information" or "does more work than he is assigned." This initiative-taking behavior was the major focus of a study attempting to identify youngsters who experienced learned helplessness (Fincham, Hokoda, & Sanders, 1989). The 12-item Pupil Behavior Checklist emphasized students' positive and negative responses when faced with obstacles to learning (e.g., 'Tries to finish assignments even when they are difficult;" "Takes little independent initiative in schoolwork; you must help him to get started on an assignment"). Correlations between total scores on the checklist and reading and mathematics performance in Grades 3 and 5 ranged from .40 to .56, all statistically significant.

The same two aspects of involvement—responding to the requirements of the class and teacher and taking an initiative with schoolwork—continue to be vital to achievement in the junior high and high school years (Anderson, 1975; Kerr, Zigmond, Schaeffer, & Brown, 1986; Laffey, 1982). In fact, the relationship of participatory behaviors and learning outcomes is so strong and consistent that Anderson and Scott (1978) proposed, "Teaching methods can be considered effective to the extent that students exhibit a high degree of involvement in learning when the methods are being used" (p. 52). This is only one criterion of effectiveness of an instructional ap-

143



Finn and Cox

proach, but it is clear that involvement in learning activities is a desirable educational outcome in its own right.

Participation in class and in school from the earliest years may also be important in reducing the likelihood of later problems including truancy or dropping out. Attendance is a form of participation that plays an important role in this connection. Although a younger child is left with little choice but to attend school and to sit in the classroom, as the youngster progresses through the grades he or she can choose to miss classes or, in the extreme, not attend school at all. Absences are confirmed as being detrimental to academic achievement and school grades (dejung & Duckworth, 1986; Weitzman et al., 1985). Even among youngsters who are in trouble with the law, those who miss the most school, i.e., truants, have the poorest school performance (Finn, Stott, & Zarichny, 1988). Moreover, Lloyd (1974, 1978) found that absences as early as Grade 6 were related to dropping out of school; absences in Grade 3 were not.

Many programs for students at risk focus on ways to increase students' involvement in school, whether in the academic, vocational, or extracurricular/social spheres. A recent overview of research on dropouts and dropout prevention programs (Wehlage, Rutter, Smith, Lesko, & Fernandez, 1989) uses the words ''participation," ''engagement," and ''involvement" 216 times in a 260-page volume. Efforts to define the constructs represented by these terms or research on the early development of participation styles, not to mention nonparticipation, are scarce despite the popularity of the terms in the literature.

Finn (1989) advances a "participation-identification model" to explain the relationship of classroom participation in the early grades with long-term behaviorial consequences. According to this formulation, most children begin school as willing participants, encouraged to become involved in classroom activities by parents and teachers. Continued participation over the years, accompanied by a degree of academic success, leads to an internalized sense of identification with school. The influence of performance rewards, coupled with the sense of belonging that identification entails, perpetuates youngsters' active involvement within and outside the classroom. The occasional frustration and less-than-successful experiences encountered by most youngsters are not sufficient to interrupt this cycle.

Unfortunately, some youngsters begin school predisposed to nonparticipation and nonidentification. Although primary grade teachers may be able to engage the interest of some of these children, others may resist participation and avoid the teacher's attention or engage in disruptive behavior. If this behavior continues over the years while academic requirements become more demanding, an accumulation of low or failing grades may be inevitable. Identification with school is unlikely to occur. The youngster may withdraw both emotionally and physically and may exhibit an expanding array of inappropriate behaviors. For a high school student whose in-

144




volvement in the social and extracurricular activities of school has become attenuated, complete withdrawal—dropping out—may seem to be a very small step. Wehlage et al. (1989) forwarded a "dropout prevention theory" based on a model with many of these same assumptions.

Research on the separate links in the participation-identification model is summarized in Finn (1989). In addition, several longitudinal studies confirm the association of early school behavior with long-term consequences. Lloyd (1978) studied over 1500 third-grade youngsters of whom 21 percent eventually left school without graduating. Dropouts and graduates were already distinct in third grade in course grades, grade retentions, and standardized achievement scores. Spivack and Cianci (1987) followed 611 inner-city school children and found significant relationships between behavior ratings in kindergarten through Grade 3 and misconduct at ages 14 and 15, school disciplinary measures, and police contacts by age 17. The more significant early ratings included "classroom disturbance," "impatience," '' disrespect-defiance," " irrelevant responsiveness,'' and '' inattentive-withdrawn."

These results and models like those presented by Finn (1989) and Wehlage et al. (1989) emphasize that behavior patterns established in the early years can persist and have long-term negative sequelae. Youngsters' participation in the classroom, school achievement, and identification with school are portrayed as interacting in a self-reinforcing cycle. Participation in class is important beginning with the child's earliest school experiences. At the same time, a degree of academic success is necessary to assure that identification will occur and that participation, rather than withdrawal, will become the youngster's habitual form of behavior.

The present study examines several aspects of this cycle, namely, the forms and antecedents of participation among fourth-grade pupils. Three groups of pupils are distinguished based on teacher ratings of their participatory and nonparticipatory behavior in the classroom: youngsters who participate actively and exhibit initiative with regard to learning activities, those who do little more than respond to the rules and instructions provided by the teacher, and those who do not participate actively and may exhibit oppositional behavior instead. The gender, race, and socioeconomic composition of these groups are examined and, finally, the groups are compared in terms of their academic achievement over the preceding 3 years.

One ancillary question is also addressed—that of the relationship of pupils' academic self-concept with classroom participation. Self-concept and self-esteem measures have low but consistent correlations with academic performance of elementary grade pupils (Byrne, 1984; Hansford & Hattie, 1982; Holly, 1987) and with later adverse behaviors such as absenteeism (Reid, 1984) and dropping out (Combs & Cooley, 1968; Bachman, O'Malley, & Johnston, 1978). Some scholars describe a chain of events whereby a youngster who receives low or failing grades may experience reduced self-

145



Finn and Cox

esteem and may stop attending class or school in an attempt to reestablish a positive self-view (e.g., Bernstein & Rulo, 1976; Wehlage & Rutter, 1986). The participation-identification model, in contrast, does not attribute a central role to the youngster's self-esteem. This investigation asks whether youngsters who participate the least in Grade 4 are those who have developed relatively poor academic self-concepts during Grades 1 through 3. Our hypothesis is that the relationship is weak or nonexistent.

Method

Subjects

Subjects for the study were 1388 fourth-grade youngsters in 72 schools across the state of Tennessee. These pupils were part of a larger group that participated in a class-size experiment beginning in kindergarten, 4 years earlier. At that time, all school systems in the state were invited to participate in the research, and about one-third of the districts representing 180 schools responded positively. After negotiation with the State Education Department, 76 elementary schools were selected; each was large enough to have at least three kindergarten classes. One-third of the kindergarten pupils in each school were randomly assigned to a small class with 12 to 17 pupils. Students were kept in the same small-class grouping through Grade 3. The other classes were maintained at the usual size, 22 to 27 pupils. The results of the investigation of class size are presented elsewhere (Finn & Achilles, 1990).

The sample for the present study is a randomly chosen subset of fourth-grade youngsters from these schools who did not attend small classes during the preceding years. The sample size was reduced further since any child who moved out of a participating school was excluded and since each teacher was asked to rate at most 10 pupils to reduce the data collection burden. A description of the sample, with schools classified as inner-city, suburban, rural, or urban, is given in Table 1. In all, about 22% of the sample is minority; of these, all but four are Black. Thirty-nine percent of the youngsters are receiving government-subsidized lunches; this provides a rough indication of socioeconomic status.

Measures

Achievement data were obtained for each student for Grades 1 through 3. Total scores were available on the Stanford Achievement Tests in reading and mathematics, administered in the spring of each year. In addition, a set of curriculum-referenced tests developed by the state of Tennessee, the Basic Skills First (BSF) tests, was administered each year. These tests are tied more closely to the State's objectives in reading and mathematics for each specific grade level (Tennessee Department of Education, 1987).

Attendance in Grade 3 was obtained by dividing the number of days

146




Table 1 Description of Fourth-Grade Sample

Location

Characteristic Inner-

city Urban Suburban Rural Total

No. of schools No. of teachers No. of students

14 42

153

5 20

101

17 61

310

36 122 824

72 245

1388

Percent of students receiving free lunch 87.6 45.8 25-3 34.9 39-4

Percent of minority students 95.4 6.9 33.9 5.5 21.8

the child was present by the number of days he or she was enrolled in the participating school. All but 16 of the youngsters in the study were enrolled for the full school year. The attendance ratio ranged from .683 to 1.000, with a median value of .97.

In November of the fourth-grade year, each student was rated by his or her teacher on the Student Participation Questionnaire. This questionnaire contains 25 items that reflect students' nonparticipatory behavior (e.g., "annoys or interferes with peers' work"), minimal adequate effort (e.g., "pays attention in class"), and initiative taking (e.g., "does more than just the assigned work"). Each item is judged according to the frequency of occurrence from (1) "never" to (5) "always." The psychometric properties of the instrument were determined for a larger sample of youngsters who participated in the class-size experiment (Finn, Folger, & Cox, 1991). Maximum likelihood factor analysis was used to identify the three factors and indicated that they could be scored separately or combined into a single composite without substantial loss of information. Coefficient alpha reliabilities for the effort, initiative, and nonparticipation scales are .94, .89, and .89, respectively.

The Self-Concept and Motivation Inventory (SCAMIN) was also administered at the end of Grades 1,2, and 3 (Milchus, Farrah, & Reitz, 1968). This group-administered instrument yields a measure of academic self-concept and academic motivation for each student; motivation scores were disregarded in the present investigation. In general, the assessment of young children's affect is difficult due to the uncrystallized, rapidly changing nature of their attitudes as well as tendency to give socially desirable responses (Goodwin & Driscoll, 1980). Within these constraints, Shepard's (1978) review notes that SCAMIN has several features to recommend it. For one,

147



Finn and Cox

a careful structured approach was taken to creating items, providing strong content and face validity. Also, the picture format in which responses are given to faces that appear happy, neutral, or sad is especially appropriate for young children. A split-half reliability of .77 was reported for the elementary-grade form. While there is an admitted lack of supportive research on construct validity, Shepard (1978) concluded that SCAMIN is useful and appropriate for making comparisons among groups and for comparing groups over time.

Procedures

Three levels of participation were identified, and each student was classified into one of the three according to his or her total score on the participation questionnaire. Ten educators were chosen to help create the three groups.1 These individuals had between 4 and 15 years of teaching experience in the elementary grades and were either currently teaching or had taught elementary school within the past 2 years. Each teacher was asked to judge the minimally acceptable level of behavior for a student on every item, using the same 5-point questionnaire scale from "never" to "always." The 10 responses to each item were combined into a single minimally acceptable level. The choice was relatively easy since the median rating of every item was also equal to its modal rating and all 10 responses were highly concentrated at or near this value. For example, the minimally acceptable level for "pays attention in class" is "usually" (4 on the 5-point scale). Both the median and mode of the 10 teachers' judgments were at this level.

Unacceptable behavior on each item was defined as one scale point below its minimally acceptable level, with appropriate reversals for negatively worded items. This is a level of behavior that would not be tolerated by the educator-judges or else is not sufficient in the judges' views for any learning to take place. For "pays attention in class," unacceptable scores were ratings of "sometimes" (3) or less. The "unacceptable" values were summed, and the nonparticipant group was defined as including all individuals whose total questionnaire scores were at or below this level.

The "minimally acceptable" values were also summed. The active-participant group was defined as including all individuals whose total scores were at or above this level. These pupils exhibited many behaviors that were acceptable or better in the eyes of their teachers. Individuals scoring between the nonparticipant and active-participant groups were considered as a third passive-participant subgroup. These individuals were in the middle —they were not outstanding in their effort or initiative in the classroom but were not sufficiently withdrawn or disruptive to receive a large number of unacceptable ratings.

The analysis of the data consisted of a comparison of the three participation groups in terms of specific classroom behaviors, demographic

148




characteristics, and the youngsters' histories of academic performance and self-concept. Multivariate analysis of variance (MANOVA) was used to compare the groups' school performance since first grade. To preserve the largest possible sample sizes, a separate MANOVA was conducted for each grade level. The dependent variables in each analysis were the four cognitive tests (SAT Reading, SAT Mathematics, BSF Reading, and BSF Mathematics). Since there are well established achievement differences between white and minority students and, to a lesser degree, between males and females, both race and sex were used as additional factors of classification in the analysis of variance design. The BSF raw score distributions were highly skewed to the left. These were transformed to x* = 10 - \n(max + 1 - x) to improve normality, where x is the original score, max the maximum total score on the test, and x* the resulting score used in the analysis.

The same 3-way analysis of variance design, in univariate form, was used to compare means on third-grade attendance ratios and self-concept for Grades 1 through 3.

When differences among the participation groups in achievement were statistically significant, two contrasts were examined—the differences of active and passive participants and the differences of nonparticipants and passive participants. The passive group was viewed as representing the middle of the behavior spectrum. These contrasts allowed us to examine patterns that were unique to highly involved pupils and to pupils who were not involved in classroom endeavors. Univariate effect sizes were estimated for each achievement test by dividing the mean difference between the participation groups by the common within-cell standard deviation of the measure; significance levels were obtained through Rests of each difference. A multivariate effect size was also obtained for each contrast. Mahalanobis's D is the number of within-cell standard deviations on a line that separates the group mean vectors or centroids. The measure is a convenient way of summarizing differences between groups when they are characterized by a set of interrelated variables (Harris, 1985, pp. 128, 168; Mahalanobis, 1930). Significance levels were obtained through Hotelling's T2 - statistic.

Since the Stanford Achievement Test scaled scores and the self-concept scores can be compared across grades, a repeated measures analysis of variance was conducted for each of these scales to see if sex, race, or participation group differences change with age. The results of this analysis are viewed as "advisory" since they involve a subset of the measures, just those pupils who had scores at all three grades, and were obtained only when the ANOVA or MANOVA suggested that effects were increasing or decreasing. The BSF scales are not comparable across grade levels.

Results

The three participation groups are compared in Table 2. Most youngsters in the nonparticipant group received unacceptable ratings on most of the

149



Finn and Cox

items in the questionnaire, and most youngsters in the active group received acceptable or higher ratings on most items. The passive group was mixed, with a typical youngster receiving acceptable ratings on slightly over half of the items. In total, the three groups were quite distinct in terms of the number of behaviors rated as acceptable.

The median rating on each questionnaire item for youngsters in each of the three participation groups was examined as well.2 All of the items showed substantial group differences.3 Responses to the most highly discriminating items of each questionnaire scale are summarized in Table 3. These items illustrate the differences in behavior that led to the classification of pupils into three groups. Passive participants are in the middle on these items and most of the others as well. These youngsters "get by." They usually complete their assignments on time, expend some energy in attempting new tasks and in completing old ones, and occasionally take a small amount of initiative with regard to learning activities.

Active participants tend to be rated at the upper extreme on each item. They are perceived as being attentive, putting significant effort into their class work when a task is difficult, and sometimes doing more work than is required. Quite in contrast, nonparticipants are seen as usually not paying attention or putting real effort in their classwork, never doing more work than required, and regularly needing to be reprimanded.

The ratings in Table 3 are group medians for separate items and do not constitute profiles for individual students. Nevertheless, the picture that they portray, especially for pupils as young as fourth-graders, is bleak. If participation in learning activities is as minimal as the ratings indicate, then

Table 2 Description of Three Participation Groups

Group

Passive- Active-Characteristic Nonparticipant Participant Participant

Total no. of pupils 226 486 676 No. of males 158 266 281 No. of females 68 220 395

Percent of minority students 29.2 22.0 19.2 Percent of students receiving

free lunch 54.8 45.5 29.7 Median number of items

rated as acceptable2 4 (16%) 13 (52%) 23 (92%) Interquartile range of number

of items rated as acceptable2 3-6 9-16 20-24

aMinimally acceptable or higher. Total number of items is 25.

150




Table 3 Median Ratings for Three Participation Groups on Highly

Discriminating Questionnaire Items

Group

Item Passive- Active-

Nonparticipant Participant Participant

Effort Scale:

Pays attention in class. Usually not

Completes homework on time. Sometimes

Approaches new assignments with sincere effort.

Tries to finish assignments even when they are difficult.

Initiative Scale:

Attempts to do his/her work thoroughly and well, rather than just trying to get by.

Does more than just the assigned work.

Goes to dictionary, encyclopedia,or other reference on his/her own to seek information. Never

Nonparticipatory Behavior Scale:

Needs to be reprimanded Usually

Sometimes Usually

Usually Always

Usually not Sometimes Always



Never Usually not Sometimes

Sometimes Usually

Sometimes Never Note. The complete 5-point scale is 1 = never, 2 = usually not, 3 = sometimes, 4 = usually, and 5 = always.

the likelihood of successful school outcomes being realized by nonpar-ticipating youngsters is poor indeed. An examination of the factors that preceded this behavior is particularly important.

The two items with the smallest differences among groups (although statistically significant) were "this student comes late to class" and "this student is withdrawn, uncommunicative." The median ratings for "comes late" were "never," "never," and "usually not," for the active, passive, and nonparticipant groups, respectively. Most fourth-grade students usually arrive at class on time, while older students may differ more on this par-

151



Finn and Cox

ticular behavior. The median ratings for "withdrawn" were "never," "usually not," and

"usually not" for the active, passive, and nonparticipant groups, respectively. These ratings indicate that even youngsters who do not participate in teacher-directed activities are not usually clinically disturbed. The classic work Psychopathological Disorders of Childhood (Quay & Werry, 1979) describes an "anxiety-withdrawal" pattern of deviant behavior as characterized by "withdrawal rather than attack . . . isolation rather than active engagement" (p. 18) and states: "If anything, this child has too little behavior rather than too much. In fact, avoidance behaviors may be so widespread that the child appears almost behaviorally paralyzed" (pp. 18-19). These phrases do not describe the pupils in the present study. Even among the nonparticipants, only 9 (4%) were rated as always being withdrawn and uncommunicative, and 31 (13.7%) were rated as usually withdrawn and uncommunicative. Most (82.3%) did not receive such extreme ratings. This stands in contrast to the percentages who usually or always "annoy or interfere with peers' work;" these were 2.8% of active participants, 11.7% of passive participants, and fully 46.9% of nonparticipants. For the most part, students in the nonparticipant group are not "behaviorally paralyzed" but fail to undertake those activities most likely to lead to positive learning outcomes.

The relationships between participation and all three background characteristics were statistically significant. For gender, x2(2, TV = 1388) = 59.09, p < .0001; for minority status, x2(2, N = 1388) = 9.89, p < .01; and for SES, x2(2, TV = 1351) = 54.92,/? < .0001. Participation levels decrease as the percent of males, percent of minorities, and percent of students receiving free lunches increase.

Differences in Past Achievement

The results of the MANOVAs are summarized in Table 4. Results for the background variables (sex and race) describe some of the general characteristics of the population being studied; mean scores are given in Table 5.4

The statistically significant difference between males and females in all three grades is attributable largely to the superior performance of males on the Stanford mathematics scale. The male advantage in mathematics was also found in the criterion-referenced test in Grade 1 but not in subsequent years. No significant difference was found between males and females on either of the reading tests at any grade level.

White students had significantly higher average performance than minorities on all cognitive measures in all three grades, with differences ranging from about 3 a to about .7o\ The effect sizes did not show a systematic increase or decrease as the youngsters progressed through the grades. The differences in performance in first grade remain throughout the subsequent 2 years.

152




Table 4 Summary of Multivariate Tests of Significance

for Achievement Measures

Grade

Effect3 1 2 3

Sex p< .0001 p < .0001 p<m Race p< .0001 p < .0001 p < .0001 Group p< .0001 p< .0001 p< .0001 Sex x race Sex x group p < . 0 5 Race x group p < . 0 5 p < . 0 1 Sex x race

x group

Sample size 1310 1323 1256

Note. Only statistically significant results indicated; results obtained from F-approximation from Wilks's likelihood ratio. aThe nonorthogonal design necessitated tests of significance in several orders (Finn & Bock, 1985). The results presented here were obtained as follows: each main effect was tested eliminating both other main effects; each 2-way interaction was tested eliminating all effects listed above it in the table; and sex x race x group was tested eliminating all else.

Multivariate tests indicate statistical significance among the three participation groups at all grade levels; mean scores are given in Table 6. In addition to the multivariate tests, every univariate F-ratio was significant at p < .0001. Two contrasts were also examined: the difference of active with passive participants and the difference of nonparticipants with passive participants. Hotelling's T2s for both contrasts were also significant a tp < .0001 at each grade level.

Multivariate effect sizes (Mahalanobis's D) for the active-passive differences are .62, .78, and .80, respectively, for Grades 1 through 3, and .45, .57, and .59, respectively, for the passive-nonparticipant differences. These indicate the number of standard deviations difference between groups for the multivariate set of measures. The achievement difference between active and passive participants was somewhat larger than the difference between passive participants and nonparticipants. Both differences increased from Grade 1 to 2 to 3. The follow-up repeated measures analysis indicated that there was no significant interaction of grade with participation groups for SAT Reading; that is, the initial first-grade differences were maintained over the two subsequent years. A significant interaction was found for mathematics; the three participation groups become more differentiated from Grade 1 to Grade 2, but there were no further increases from Grade 2 to Grade 3. Overall, the three participation groups had clearly distinct school

153



Table 5 a Mean Outcome Scores by Sex and Race a

a. Sex Race

& Effect Effect

Variable Grade Male Female size3 White Minority size3

SAT Reading 1 531.3 537.5 .01 541.8 508.3 .66*** 2 587.5 595.7 - . 0 3 598.2 567.5 .74*** 3 616.0 623.0 - . 0 3 623.5 605.8 .48***

SAT Mathematics 1 541.9 540.5 .19*** 546.3 523.5 .58*** 2 588.6 587.9 .20*** 594.1 567.8 .61***

3 622.1 623.9 .14* 627.6 607.2 .51***

BSF Reading13 r 8.63 8.77 - . 0 5 8.76 8.49 .28*** 2 8.10 8.27 - . 0 4 8.29 7.81 .59*** 3 8.11 8.24 - . 0 4 8.24 7.95 .38***

BSF Mathematics13 1 8.92 8.85 .20*** 8.99 8.54 .50*** 2 8.29 8.36 .08 8.42 7.97 .52*** 3 8.05 8.20 .00 8.25 7.70 .62***

Note. Significance of univariate F-statistic as follows: *p < .05; **p < .01; ***p < .001. aLeast squares estimate of mean difference, divided by common within-group standard deviation, transformed values as described in text.




Table 6 Mean Achievement Scores by Participation Group

Variable

Participation SAT SAT BSF BSF Grade group Reading Mathematics Reading Mathematics

Nonparticipant Passive Active

505.6 521.9 552.6

521.2 532.2 554.1

8.30 8.53 8.95

8.50 8.76 9.11


562.6 579.2 609.7

561.1 577.4 604.9

7.63 7.96 8.52

7.77 8.16 8.62


594.1 608.4 635.2

596.0 611.8 639.2

7.70 8.00 8.44

7.50 7.87 8.50

Common within-group standard deviations:2

Grade 1 45.61 34.85 0.79 0.79 Grade 2 37.73 38.36 0.71 0.78 Grade 3 31.59 34.52 0.66 0.79

aPooled across all 12 cells in the sex x race x group design.

performance levels from first grade onward. Interactions. The multivariate test of the race- x -group interaction was

significant at p < .05 in Grade 2 and at p < .01 in Grade 3; although the multivariate test was not significant in Grade 1, the four univariate F-ratios had p-values between .06 and .09. Multivariate tests of particular interaction contrasts at each grade level revealed that the difference active-passive interacts with race while the difference nonparticipant-passive does not.

To illuminate further, differences among the participation groups on each scale were estimated separately for White and minority students (see Table 7). The performance difference between active and passive participants was larger for White students on all four tests at all three grade levels. The means for separate race groups (not tabled) show that passive participants had the smallest race difference in performance. The race difference was significantly larger among active participants on each test. That is, White students who take initiative in learning activities benefit more in reading and mathematics than their Black peers who display the same behavior. The reasons for this are not clear.

Table 7 indicates that nonparticipants scored significantly below passive participants on every test at each grade level. The difference ranged from approximately one-fourth to two-thirds of a standard deviation. While these

155



Finn and Cox

decrements in achievement were generally larger for minorities than for White students, the interaction of race with the nonparticipant-passive differences was not statistically significant. In other words, the handicap associated with nonparticipation in learning activities is just as debilitating for White as for Black pupils. Further, the absence of a significant sex- x -race- x -group interaction indicates that the nonparticipation handicap is debilitating for boys and girls alike.

The significant multivariate test of sex- x -group interaction in Grade 3 is attributable entirely to the BSF Mathematics scale, on which there was

Table 7 Differences Among Participation Groups for

White and Minority Students

Contrast

Grade

Measure Contrast 1 2 3

SAT Reading: Active-Passive White .72*** .86*** .95*** Minority .39** .48*** .42**

Non-Passive White - .30** - .39*** - .35*** Minority - . 3 9 * - . 45** - . 63***

SAT Mathematics: Active-Passive White .71*** .82*** .87*** Minority .34** .34* .55***

Non-Passive White - . 2 7 * * - .37*** - .37*** Minority - . 4 3 * * - . 63*** - .67***

BSF Reading: Active-Passive White .59*** .79*** .71*** Minority .26 .68*** .45***

Non-Passive White - . 2 1 * - .40*** - .35*** Minority - . 4 5 * * - . 50** - .62***

BSF Mathematics: Active-Passive White .53*** .64*** .88*** Minority .19 .34* .43**

Non-Passive White - . 2 7 * * - .50*** - .42*** Minority - .50** - . 49** - .46**

Note. All values are least squares estimates of mean differences divided by the common within-group standard deviation. Statistical significance is indicated as follows: *p < .05; **p < .01; ***p< .001.

156




a gender reversal among the active participants. While boys outperformed girls in the nonparticipant and passive groups and in all groups in Grade 1 and 2, actively participating girls had higher average performance in Grade 3 than their male peers. This reversal was not found for any of the other measures.

The role of attendance. Attendance in the early grades is not a matter of pupil discretion as it may become in later years. This fact is reflected in a high mean attendance ratio for third-grade pupils—.96, with standard deviation .036. Nevertheless, at all ages attendance may serve as a necessary but not sufficient condition for other forms of participation to occur. As the youngster's autonomy increases with age, the decision to attend school or a particular class becomes a highly significant participatory behavior.

The same analysis of variance model used with the achievement scores was applied to the attendance measure. Although the range of attendance ratios was limited, statistically significant differences were found for sex [F(l,1244) = 10.10,p< .01], for race [F(l, 1244) = 7.30, p < .01] and for participation groups [F(2, 1240) = 10.05, p < .0001]. On average, males had slightly better attendance records than females (.963 compared with .958), and minorities had slightly better attendance than White pupils (.966 compared with .960). The mean attendance rates for participation groups were .965, .959, and .955 for active, passive, and nonparticipants, respectively. Tests of individual contrasts indicated that the active-passive difference was significant a tp < .01, while the passive-nonparticipant difference was significant 2ip < .05 if a one-tail test is made. No interaction was statistically significant; different attendance rates among participation groups were found among males and females and among Whites and minorities alike. Again all of the differences are small in absolute terms, but attendance is monoton-ically related to participation and the differences are statistically reliable. This may be all that can be expected of this measure in Grade 3.

Group Differences in Self-Concept

The 3-way ANOVA for SCAMIN self-concept scores in first grade yielded no significant differences between sex groups, race groups, or participation groups and no significant interactions. In second grade, the average self-concept score for females was significantly higher than that for males by .16a[F(l, 1293) = 8.37,p < .01] and the mean for minorities was significantly higher than that for Whites by .20a [F(l, 1293) = 9 .19 ,p< .01)]. No other statistically significant effects were obtained. In third grade, females maintained a higher average self-concept score by .12CT[F(1, 1345) = 4.69, p < .05] and minorities by .23cr [F(l, 1345) = 12.36, p < .001]. The pattern of results for minorities—higher average self-concept but lower average achievement—has been found in other studies (Wylie, 1979). While the reasons for the apparent inconsistency are unclear, poorer performance on achievement tests neither causes nor results from low average self-concept

157



Finn and Cox

for minority youngsters. Differences among the participation groups were statistically signifi

cant in Grade 3 [F[2, 1345) = 4.13, p < .05]. Of the two contrasts examined, active participants had a higher average self-concept than passive participants by .l6a. Nonparticipants were not significantly different from the passive group. There were no significant interactions with participation.

In sum, there were no significant differences among groups on the self-concept measure in Grade 1, significant sex and race differences in Grade 2, and significant sex, race, and participation group differences in Grade 3. The youngsters in this investigation appeared to become increasingly differentiated with age. The follow-up repeated measures analysis confirmed that there was a significant interaction of grade with sex a tp < .01 and with race a tp < .001, but not with participation. The differences among participation groups in Grade 3 were not significantly greater than the (nonsignificant) differences among the same groups in Grades 1 and 2. It may be that group differences in self-concept are just emerging among third-grade pupils and will become accentuated in later grades; or it may be that an effect was observed in this particular sample but will not replicate.

In general, when active participants, passive participants, and nonparticipants in fourth grade are compared on average self-concept over the preceding 3 years, few if any differences are found. This finding is noteworthy. It is not the students with the poorest self-view who withdraw from participation in the classroom, and it is not necessary to enhance a youngster's self-concept in order for him or her to become engaged in learning activities. Further, the correlations of self-concept with the four achievement tests over the preceding years range only from - .047 to + .069. While participation is closely related to achievement in the first three grades, a youngster's academic self-concept is less relevant to either of these, if at all.

Discussion

This study examined the relationship between participation in learning activities and past school achievement for a sample of 1388 fourth-grade pupils. Based on their teachers' ratings, the youngsters were classified into three participation groups: active participants, passive participants, and non-participants. The rating scale assessed three aspects of participation: the degree to which students expend effort to meet the minimal requisites of the teacher and class setting, the degree to which students display initiative with regard to learning, and the degree to which youngsters exhibit non-participatory or disruptive behavior. While the groups were formed on the basis of total participation scores, they were also distinct in terms of the ratings received on individual items. The inadequate effort level of non-participants, accompanied by the display of oppositional behavior, was particularly salient.

When the three participation groups were compared in terms of achieve-

158




ment in Grades 1 through 3, there were highly significant differences at all three grade levels, with multivariate effect sizes ranging from .45 to .80. The differences among the groups were clearly distinct even in first grade and persisted over the ensuing years. Other researchers have demonstrated that the degree of participatory behavior (e.g., attention, attendance) seen early in students' careers differentiates successful from unsuccessful students in later years (Attwell et al., 1967; Lloyd, 1978). The present findings add support to the argument for the early identification of students with inadequate levels of participation.

There was no significant interaction of the contrast nonparticipant-passive participant with race. The decrement in achievement due to withdrawing from class activities is a significant problem for Whites and minorities alike. At the same time, the nonparticipant group had a higher percentage of minority students than the other groups, and the average school performance of minorities was lower than the performance of Whites at all grades. It is unfortunate that this investigation did not have participation measures for earlier grades. Nevertheless, the data strongly suggest that efforts to increase students' participation behaviors in the first years of school could benefit minorities in particular.

This proposition receives further support from attribution theory. It is fairly well established that young children attribute success outcomes to effort, while older children develop a concept of ability as an explanation for performance differences; the shift often occurs around age 9 (Nicholls, 1978, 1979, 1989). If a youngster does not become involved in learning activities in the first two or three grades or enjoy the successful experiences that participation can bring, it is highly unlikely that he or she will become involved in later grades when an ability requirement is perceived as an additional obstacle to learning. Youngsters who exhibit nonparticipation and inadequate performance in the early years may engage in "failure-avoiding" or "failure-accepting" behavior in later years (Covington & Omelich, 1985). A major component of failure-accepting behavior is further reduction in the expenditure of effort, that is, continued nonparticipation.

In general, participation groups did not differ significantly on the measure of academic self-concept; nor was there any significant interaction of sex or race with participation groups in self-concept. This is consistent with Finn's (1989) argument that self-concept is a not a prominent mediator of school performance, at least in the early grades. According to the participation-identification model, participation in the classroom in the younger years, especially if followed by successful outcomes, leads eventually to a sense of identification with school. That is, the youngster feels that he or she "belongs" in the school setting and comes to value learning outcomes. This positive affect, in turn, serves to perpetuate the youngster's participation in learning-relevant activities. It may be difficult or futile to attempt to increase youngsters' self-concepts in the hope that achievement

159



Finn and Cox

will improve. Instead, every effort should be made to identify nonpartici-pating students in the youngest years and to encourage them to become involved in learning activities, over and over again. If some degree of success follows, participation will be reinforced as a mode of behavior and the cycle will become self-sustaining.

Further research is needed to increase our understanding of different forms of participation in school and the classroom, that is, to give further definition to constructs such as "involvement" and "engagement." An increased understanding of the antecedents and consequences of nonparticipa-tion is particularly important. This study has explored participation among elementary grade students. The assessment of participation encompassed a limited but important set of behaviors that are exhibited by young school children. As students mature through the grades, there are increased opportunities to become involved in the social, extracurricular, and even decision-making sectors. For a student who is predisposed to withdraw from participation in school, all of these have the potential for maintaining his or her involvement and reducing the likelihood of complete withdrawal. We should seek to identify those institutional practices that promote involvement in schooling in both its basic and elaborated forms.

Notes

This work was sponsored by a grant from the Spencer Foundation. The authors are grateful to Kristin Voelkl for her assistance with reference materials and to Tennessee's Project STAR directors and staff for collecting the data for this investigation. Appreciation is extended to Elizabeth Word, project director; Helen Bain, John Folger, John Johnston, Nan Lintz, and Charles Achilles, the STAR Consortium; Barbara Nye and the Center for Excellence at Tennessee State University; and DeWayne Fulton and Jayne Zaharias, research associates.

This article was completed while the first author was American Statistical Association Fellow at the National Center for Education Statistics.

'These teachers were totally distinct from those who rated the fourth-grade pupils. 2The median and modal ratings were identical in almost all instances. 3A x2-test of association between the participation groups and the responses to each

item was computed. Although the items were used in forming the participation groups initially, this gave an adequate ranking of the discriminating power of the items. All x2-values were highly significant, as might be expected for 1388 observations; the smallest x2 out of 28 was x2 (8) = 200.0.

4The astute reader will note that four of the sex differences in Table 5 have an apparent inconsistency. The female means for SAT Reading (Grade 1), for SAT Mathematics (Grade 3), and for BSF Mathematics (Grades 2 and 3) are higher than the male means, while the effect size is in the opposite direction. This occurs because of the impact of the unequal ATs in the 3-way design. The sample means are the simple averages of all males and all females, respectively, regardless of where the individuals fall with respect to race or participation group. Classifications on these other two dimensions are said to confound the means for sex groups. The analysis of variance model corrects for this source of bias, and obtains simultaneous estimates of sex, race, and participation differences given the arbitrary distribution of subjects in the design. The correct least-squares estimate of the population difference "male-effect-minus-female-effect" on SAT Reading in Grade 1 is .01 (standard deviation). The analysis of variance model can be used to derive estimates of population means for males and females that have exactly this difference, but the explanation is complex and may be unnecessary. Fortunately, the remaining effect sizes in Table 5 are all in the same direction as the difference between the sample means.

160




References

Anderson, L.W. (1975). Student involvement in learning and school achievement. California Journal of Educational Research, 26, 53-62.

Anderson, L.W., & Scott, C.C. (1978). The relationship among teaching methods, student characteristics, and student involvement in learning. Journal of Teacher Education, 29(3), 52-57.

Attwell, A.A., Orpet, R.E., & Meyers, C.E. (1967). Kindergarten behavior ratings as predictors of academic achievement. Journal of School Psychology, 6(1), 43-46.

Bachman, J.G., O'Malley, P.M., & Johnston, J. (1978). Adolescence to adulthood: Changes and stability in the lives of young men. Ann Arbor: University of Michigan, Institute for Social Research.

Bernstein, S., & Rulo, J.H. (1976). Learning disabilities and learning problems: Their implications for the juvenile justice system. Juvenile Justice, 27 (4), 43-47.

Byrne, B.M. (1984). The general/academic self-concept nomological network: A review of construct validation research. Review of Educational Research, 54, 427-456.

Cobb, J. A. (1972). Relationship of discrete classroom behaviors to fourth-grade academic achievement. Journal of Educational Psychology, 63, 74-80.

Combs, J., & Cooley, W.W. (1968). Dropouts: In high school and after school. American Educational Research Journal, 5, 343-363-

Covington, M.V., & Omelich, C.L. (1985). Ability and effort valuation among failure-avoiding and failure-accepting students. Journal of Educational Psychology, 77, 446-459.

dejung, J.E., & Duckworth, K. (1986, April). Measuring student absences in the high schools. Paper presented at the annual meeting of the American Educational Research Association, San Francisco.

Dewey, J. (1929). My pedagogic creed. Journal of the National Education Association, 18 (9), 291-295.

Fincham, F.D., Hokoda, A., & Sanders, Jr., R. (1989). Learned helplessness, test anxiety, and academic achievement: A longitudinal analysis. Child Development, 60, 138-145.

Finn, J.D. (1989). Withdrawing from school. Review of Educational Research, 59, 117-142.

Finn, J.D., & Achilles, CM. (1990). Answers and questions about class size: A statewide experiment. American Educational Research Journal, 27, 551-511.

Finn, J.D., & Bock, R.D. (1985). MULTIVARIANCE VII User's Guide. Chicago: Scientific Software.

Finn, J.D., Folger, J., & Cox, D. (1991). Measuring participation among elementary grade students. Educational and Psychological Measurement, 51, 393-402.

Finn, J.D., Stott, M.W., & Zarichny, K. (1988). Low-achieving students in juvenile court. Urban Education, 23, 150-161.

Goodwin, W.L., & Driscoll, L.A. (1980). Handbook for measurement and evaluation in early childhool education. San Francisco: Jossey Bass.

Hansford, B.C., & Hattie, J.A. (1982). The relationship between self and achievement/performance measures. Review of Educational Research, 52, 123-142.

Harris, R.J. (1985). A primer of multivariate statistics (2nd ed.). Orlando: Academic Press.

Holly, W.J. (1987). Student self-esteem and academic success [Whole issue]. Oregon School Study Council Bulletin, 31 (2).

Kerr, M.M., Zigmond, N., Schaeffer, A.L., & Brown, G.M. (1986). An observational follow-up study of successful and unsuccessful high school students. High School Journal, 70, 20-24.

161



Finn and Cox

Laffey, J.M. (1982). The assessment of involvement with school work among urban high school students. Journal of Educational Psychology, 74, 62-71.

Lahaderne, H.M. (1968). Attitudinal and intellectual correlates of achievement: A study of four classrooms. Journal of Educational Psychology, 67, 198-203.

Lloyd, D.N. (1974). Analysis of sixth-grade characteristics predicting high school dropout or graduation. JSAS Catalog of Selected Documents in Psychology, 4, 90.

Lloyd, D.N. (1978). Prediction of school failure from third-grade data. Educational and Psychological Measurement, 38, 1193-1200.

Mahalanobis, P.C. (1930). On tests and measures of group divergence. Journal of Asiatic Society of Bengal, 26, 541-588.

McKinney, J.D., Mason, J., Perkerson, K., & Clifford, M. (1975). Relationship between classroom behavior and academic achievement. Journal of Educational Psychology, 67, 198-203-

Milchus, N., Farrah, G., & Reitz, W. (1968). The Self-Concept and Motivation Inventory: What face would you wear? Dearborne Heights, MI: Person-O-Metrics.

Nicholls, J.G. (1978). The development of the concepts of effort and ability, perception of own attainment, and the understanding that difficult tasks require more ability. Child Development, 49, 800-814.

Nicholls, J.G. (1979). Development of perception of own attainment and causal attributions for success and failure in reading. Journal of Educational Psychology, 71, 94-99.

Nicholls, J.G. (1989). The competitive ethos and democratic education. Cambridge: Harvard University Press.

Quay, H.C., & Werry, J.S. (1979). Psychopathological disorders of childhood (2nd ed.). New York: Wiley.

Reid, K.C. (1984). Some social, psychological, and educational aspects related to persistent school absenteeism. Research in Education, 31, 63-82.

Shepard, L.A. (1978). Review of The Self-Concept and Motivation Inventory (SCAMIN): What face would you wear? In O.K. Buros (Ed.), The eighth mental measurements yearbook (Vol. 1, pp. 1053-1054). Highland Park, NJ: Gryphon Press.

Spivack, G., & Cianci, N. (1987). High-risk early behavior pattern and later delinquency. In J.D. Burchard & S.N. Burchard (Eds.), Prevention of delinquent behavior (pp. 44-74). Beverly Hills, CA: Sage Publications.

Swift, M.S., & Spivack, G. (1969). Achievement-related classroom behavior of secondary school normal and disturbed students. Exceptional Children, 35, 677-684.

Tennessee Department of Education. (1987). STAR Criterion Referenced Test. Manual for Test Administration. Grade 1. Nashville: State of Tennessee.

Wehlage, G.G., & Rutter, R.A. (1986). Dropping out: How much do schools contribute to the problem? Teachers College Record, 87, 374-392.

Wehlage, G.G., Rutter, R.A., Smith, G.A., Lesko, N., & Fernandez, R.R. (1989). Reducing the risk. Schools as communities of support. London: Falmer Press.

Weitzman, M., Klerman, L.V., Lamb, G.A., Kane, K., Geromini, K.R., Kayne, R. Rose, L., & Alpert, J J. (1985). Demographic and educational characteristics of inner city middle school problem absence students. American Journal of Orthopsychiatry, 55, 378-383.

Wylie, R.C. (1979). The self-concept (revised ed.). Lincoln: University of Nebraska Press.

162

