You are on page 1of 17

American Educational Research Journal

Spring 1992, Vol. 29, No 1, pp. 182-198

Kindergarten Retention: Academic and


Behavioral Outcomes Through the
End of Second Grade
Panayota Mantzicopoulos
Purdue University
and
Delmont Morrison
University of California, San Francisco

This study examined the impact of retention at kindergarten on academic


achievement and behavior. Subjects were 53 children, retained at kin-
dergarten, who were matched to a group of 53 promoted peers on demo-
graphic characteristics, a measure of school readiness, and preacademic
achievement in reading and mathematics. The data were analyzed using
both same-age and same-grade comparisons. Results indicated an aca-
demic advantage of the retained children during their second year in
kindergarten. This advantage was not maintained past kindergarten.
Although retained children demonstrated a decline in attention problems
during their second year of kindergarten, they continued to perform below
the norm for their school districts on academic achievement. The data
in this report do not suggest that retention is an effective policy for the
young at-risk child.

PANAYOTA MANTZICOPOULOS is an assistant professor in the School of Educa-


tion at Purdue University, Educational Psychology, SCC-G, West Lafayette, IN 47907.
She specializes in research on students at risk for learning and behavior problems
and on students' coping with school failure.
DELMONT MORRISON is a clinical professor in the Department of Psychiatry at
the University of California-San Francisco, 401 Parnassus Ave., San Francisco, CA
94143. He specializes in child clinical psychiatry, learning disabilities, and at-risk
factors in infants and preschool children.

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention

R etention rates are on the rise despite research demonstrating that the
practice is not effective (Holmes & Matthews, 1984; Jackson, 1975;
Smith & Shepard, 1987). Unlike mixed empirical evidence on other educa-
tional issues, research on elementary school nonpromotion is unequivocal.
It supports the conclusion that retention is not an effective policy. There
seem to be clear indications that the practice not only fails to remediate
children's academic problems, but is also associated with negative self-
concepts in children, negative attitudes toward school, and higher dropout
rates (Byrnes & Yamamoto, 1986; Grissom & Shepard, 1989).
In recent years, attention has shifted to grade repetition at the kinder-
garten level, and the controversy regarding the usefulness of this practice
is slowly escalating. At the heart of the matter are notions concerning the
importance of (a) developmental readiness and (b) the acquisition of basic
academic skills as early as possible. Thus, kindergarten retention is often
the intervention of choice for children who are socially immature or have
difficulty acquiring basic academic skills.
Proponents of early retention assert that the practice is substantially
more beneficial to children than retention in the elementary grades. In fact,
these educators argue that retention at kindergarten is a different case from
later retention for two reasons. First, the practice is aimed at prevention.
Second, since retention comes before academic failure, it does not carry
the stigma attached to retention in later grades (Shepard, 1989).
This line of reasoning has led to the early application of formal and
informal assessment procedures seeking to identify children who are not
ready to profit from the first grade curriculum. When these children are
identified, the school may recommend that they spend an extra year in kin-
dergarten. The underlying assumption is that an extra year in the same en-
vironment will help children reach the maturational level necessary for a
successful first grade experience.
This assumption can be called into question when one considers re-
cent research on kindergarten retention. One study, for example, indicated
that children retained at the end of kindergarten performed lower than a
random sample of their promoted peers on a number of measures (Mantzi-
copoulos, Morrison, Hinshaw, & Carte, 1989). Specifically, at the end of
their first year in kindergarten, children recommended for retention had
lower IQ and preacademic achievement test scores. In addition, they ex-
hibited more problems in perceptual organization, in visual-motor integra-
tion, and in the areas of attention, impulsivity, and hyperactivity. The
authors thus argued that this evidence was ''incompatible with the assump-
tion that reexposure to the same curriculum is enough to facilitate remedia-
tion [of the retained group's deficits]" (p. 119).
Further support for this view comes from reviews of available kinder-
garten retention and transition room studies (Shepard, 1990; Shepard &
Smith, 1986). These reviews concluded that both practices are ineffective.

183

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
As Shepard and Smith (1986) state, 'There is no achievement benefit in
retaining a child in kindergarten or first grade and, regardless of how well
the extra year is presented, the child still pays an emotional cost" (p. 40).
In the most recent review by Shepard (1990), only one well-controlled study
was identified where an academic advantage was found for kindergarten
retained children. However, it is not known whether this advantage was
maintained, because data were not provided beyond first grade (Shepard,
1990).
The scarcity of studies with longitudinal information, adequate con-
trol groups, and simultaneous investigations of social/emotional and aca-
demic outcomes limits our ability to draw conclusions when positive ef-
fects are reported. In addition, another point of concern is that retained
children are a year older than their peers when comparisons are made. This
age difference may bias results in favor of children who have been retained.
A useful way of addressing this issue was introduced in the design of
a recent longitudinal study of elementary school nonpromotion (Peterson,
DeGracie, & Ayabe, 1987). Under this design, retained students are com-
pared to their matched counterparts in two ways (a) within the same year
(or same-age group) and (b) within the same grade. For example, in same-
year comparisons, retained children in their second year of kindergarten
are compared to matched promoted peers in first grade. In this case, both
groups are the same age but the matched promoted peers are taking a dif-
ferent level of the test. In same-grade comparisons, retained and promoted
children's scores are compared at the end of each grade. In this case, re-
tained children are a year older than their matched counterparts.
This design was adopted by Banerji (1990) in her 4-year investigation
of children in a 2-year developmental kindergarten. This study examined
academic effects using both same-age and same-grade comparisons of re-
tained children and matched promoted controls. Based on her analyses,
Banerji (1990) concluded that although there was no evidence of bene-
fit for the transitional children there was also no evidence to show that
placement in a developmental kindergarten harms immature children
academically.
The present study employs the same design and explores both academic
and behavioral effects of kindergarten retention with a sample of retained
and promoted children followed until the end of second grade.
Method

Sample
The sample for this study came from two school districts in Marin County,
California. Both districts were participating in a large longitudinal project
on the effectiveness of the SEARCH screening instrument (Silver & Hagin,
1981). The data for this study came from two cohorts of children. The first

184

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
cohort included 35 children who were retained during the academic year
1985-86. The second cohort included 18 children who were retained in
1986-87. The final group of 53 retained children had complete data on
academic achievement tests from kindergarten to second grade.
Retained children were matched at kindergarten to 53 promoted peers
from the larger pool of subjects who were participating in the SEARCH
study. Matching was accomplished on a case by case basis. Matching vari-
ables were cohort, school, sex, age, at-risk status (as measured by SEARCH),
reading achievement, and math achievement. Children were also matched
on socioeconomic status when data were available. SES data were available
on 87 (82.1%) children (36 promoted and 51 retained) only.
When the matched pairs were created, there was strict adherence to
the requirement that children be of the same cohort, school, and sex. To
keep this requirement, small deviations in the other variables were allowed
when an identical match could not be found for a retained child. These
deviations were (a) 2-3 months for age, (b) a difference of one socioeco-
nomic status level, and (c) a difference of .20 standard deviations in achieve-
ment test scores. Because of the limited predictive validity of the SEARCH
screening instrument (Morrison & Mantzicopoulos, 1990), we allowed some
matched pairs to deviate from their risk status designation (1 = at-risk,
0 = not-at-risk) when a good match for a child was obtained on all other
variables. The final data on the matching variables are presented in Table
1. A series of f-tests were conducted to ensure that retained students were
effectively matched to their promoted peers. None of the comparisons was
significant, indicating that the two groups were equivalent on the matching
variables.
To summarize, the age range of the sample at the time of the SEARCH
administration was 63 to 74 (M = 66.06) months. There were 38 males and
15 females in each group. The racial composition of the sample was as fol-

Table 1
Means and Standard Deviations of Retained and Promoted
Groups on Matching Variables

Retained2 Promoted2
Variable M SD M SD

Age in months 65.51 3.01 66.62 3.45


SES 3.44 0.99 3.14 1.08
At-risk status 0.71 0.46 0.51 0.51
Total reading 5 -0.38 0.69 -0.36 0.57
Total math b -0.45 0.64 -0.35 0.54

a
N = 53; Standard (z) scores

185

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
lows: 83 (78.3%) Caucasian and 23 (21.7%) Black. Of the 87 (82.1 %) chil-
dren on whom socioeconomic status data were available, 4 (4.6%) were
in Class I, professional; 15 (17.2%) were in Class II, upper middle; 37 (42.5%)
were in Class III, middle; 23 (26.4%) were in Class IV, lower middle; and
8 (9.2%) were in Class V, unskilled labor and unemployed.
Behavioral subsample. Teacher ratings of behavior (assessed by the
Revised Behavior Problem Checklist described later) were available only
on a subsample of 32 retained and 30 promoted children. This was because
of the constraints imposed by the larger SEARCH study from which the
data for this investigation were obtained. The larger study was designed
to examine the characteristics of two groups of children: a group designated
"at-risk" on SEARCH and a matched not-at-risk control sample (Morrison,
Mantzicopoulos, & Carte, 1989). Information on academic achievement was
available for all children. However, teacher behavior ratings were not ob-
tained for those not-at-risk children who did not serve as not-at-risk con-
trol subjects in the larger SEARCH study. This procedure accounted for the
smaller subsample of children (n = 62) who served as subjects for the report
on the behavioral effects of retention.
This subsample, compared to the entire sample, consisted of higher
proportions of children designated at-risk by SEARCH. Specifically, the orig-
inal sample (TV = 106) included 60 at-risk (26 promoted and 34 retained)
and 46 not-at-risk (27 promoted and 19 retained) children. By contrast, the
subsample included 50 at-risk (23 promoted and 27 retained) and only 12
not-at-risk (7 promoted and 5 retained) children. This initial difference in
the at-risk status of the two samples made it necessary to reexamine devia-
tions on the matching variables of the study between (a) the subsample and
the larger sample and (b) the remaining 32 retained and 30 promoted chil-
dren who would serve as subjects for the behavioral comparisons. Results
from these comparisons are presented in Table 2. First, f-tests were con-
ducted within each group to investigate differences between children in
the original sample and the remaining subsample on all the matching
variables. None of the comparisons reached significance, suggesting that
the scores of children in the subsample did not deviate significantly from
those in the original sample. Second, f-tests were used to examine differences
between children in the retained (n = 32) and promoted (n = 30) groups
of the subsample. These comparisons were not significant, indicating that
the two groups remained equated on the matching variables of the study.
The gender ratio of this sample was 19 males to 11 females for the
promoted group and 24 males to 8 females for the retained group. We were,
therefore, concerned about possible biases resulting from comparing groups
with a male/female ratio of 3/1 versus 2/1. For this reason, the data were
converted to standard (z) scores using separate means and standard devia-
tions for males and females at each grade level. The means and standard
deviations of the normative sample of regular school children, reported in

186

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
Table 2
Means and Standard Deviations of Children in Original
Sample and Behavioral Subsample

Retained Promoted
Variable M SD M SD

Age in months
Sample 65.51 3.01 66.62 3.45
Subsample 66.00 3.31 66.67 3.48
SES
Sample 3.44 0.99 3.14 1.08
Subsample 3.42 0.96 3.00 1.07
At-risk status
Sample 0.71 0.46 0.51 0.51
Subsample 0.84 0.36 0.77 0.43
Total reading 3
Sample -0.38 0.69 -0.36 0.57
Subsample -0.52 0.76 -0.30 0.61
Total math 3
Sample -0.45 0.64 -0.35 0.54
Subsample -0.42 0.80 -0.42 0.66

Standard (z) scores

the manual of the Revised Behavior Problem Checklist (Quay & Peterson,
1987), were used for the z-score conversion.
Measures
SEARCH. This individually administered instrument is comprised of
10 subscales that assess the spatial and temporal orientation believed basic
to beginning reading (Silver & Hagin, 1981). This instrument is designed
for children between 63 and 80 months of age. The individual administra-
tion takes approximately 20 minutes and is to be given prior to entry into
first grade. The ten subtests measure visual discrimination, immediate visual
recall, visual-motor copying, rote sequencing, auditory discrimination, artic-
ulation, associating sounds with their visual symbols, directional orienta-
tion, finger schema, and pencil grip. Internal consistency coefficients for
the ten subtests were reported to range from 0.36 to 0.93 with a mean of
0.69 (Silver, Hagin, DeVito, Kreeger, & Scully, 1976). Test-retest reliability
over a 2-week interval was 0.71 (Silver et al., 1976). A recent analysis of
the predictive validity of SEARCH with a group of Northern California
children (N = 668) from which the sample of this study was drawn is
reported in Morrison and Mantzicopoulos (1990). That study indicated that
187

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
SEARCH has, on the average, a 74% rate of correct classifications and a
correlation of .23 with achievement at the end of second grade.
The procedures for administration and scoring suggested by Silver and
Hagin (1981) were followed and local norms were used. Thus, as recom-
mended by the authors of SEARCH, children scoring below the 33rd percen-
tile locally were considered at-risk for reading failure. It should be men-
tioned that with this population the lack of an appropriate pencil grip did
not discriminate between vulnerable and nonvulnerable children. This is
because 48% of the children in the Marin County sample demonstrated
an appropriate pencil grip. This is in contrast to the 75-79% reported for
the original SEARCH sample (Silver & Hagin, 1981). For this reason, SEARCH
was scored from 0 to 9. Scores of 5 or less indicated at-risk performance
and were assigned the code 1. Scores higher than 5 indicated not-at-risk
performance and were assigned the code 0.
Socioeconomic status (SES). SES was assessed using Hollingshead's
(1957) Two-Factor Index of Social Position, asking parents to provide in-
formation on their occupation and highest educational level attained.
Revised Behavior Problem Checklist (RBPC). Constructed by Quay and
Peterson (1987), this 89-item checklist provides six subscales: Conduct
Disorder, Anxiety Withdrawal, Attention Problems-Immaturity, Motor Ex-
cess, Socialized Aggression, and Psychotic Behavior. Behaviors are scored
on a 3-point scale where 0 reflects not a problem, 1 indicates a mild prob-
lem, and 2 represents a severe problem. Because of the age range of the
children in this study, the Socialized Aggression scale (steals in company
with others, loyal to delinquent friends, uses alcohol in the company of
others, etc.) was inappropriate and was not used. The factorial validity of
the RBPC has been established with at least six different samples of children
(Quay & Peterson, 1987). Alpha reliabilities computed from different samples
of school children range from .68 to .94. Test-retest reliabilities over a
2-month interval range from .61 for Psychotic Behavior to .83 for Atten-
tion Problems. The scale provides norms by gender and grade, based on
a sample of "869 unselected public school children in suburban and rural
schools" (Quay & Peterson, 1987, p. 12). These normative data were used
to convert each child's raw score on each subscale to a standard score.
Group academic achievement tests. Depending on their district, chil-
dren were tested with either the Stanford Achievement Test (Gardner, Mad-
den, Rudman, Karlsen, Merwin, Callis, & Collins, 1985) or the California
Test of Basic Skills (CTBS, 1982). From each test, the Total Reading and
Total Math subtests were used for the analyses. In order to ensure com-
parability of scores between tests, scale scores were converted to standard
(z) scores using national norms provided in the technical manuals of each
test. Because children in Marin County, on the average, perform above the
mean on group academic achievement tests, a z score conversion was also
made using local norms. The means and standard deviations for each test
within each district were used for this conversion.
188

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
Procedure
The SEARCH screening instrument was individually administered to the sub-
jects during the month of February of their kindergarten year. In March
of kindergarten, a questionnaire was mailed to parents assessing parental
education and occupation. In April of each year, teachers rated each child
on the RBPC. Academic achievement tests were administered by the schools
in the month of May of each year.
It should be noted that SEARCH was not an adopted screening device
by the public schools. The schools participating in the present investiga-
tion were not involved in early screening practices. The administration of
SEARCH was part of an independent research effort directed at establishing
the test's validity for early screening purposes.
Teachers who provided information on each student knew that the
majority of their students were part of a large screening study. However,
they were not aware that retained students were also the target of the in-
vestigation (78% of the children in Marin County schools participated in
the screening study investigating the predictive validity of SEARCH). Chil-
dren's scores on each measure were not reported to classroom teachers.
Teachers were only aware of children's scores on the group academic
achievement test administered by the school. The RPBC was not used in
retention decisions. Schools made retention recommendations by consider-
ing children's academic achievement scores as well as their level of matur-
ity, as judged by the classroom teacher.
This study did not evaluate the similarity of instructional practices across
kindergarten programs. However, the participating kindergartens were con-
sidered, in these districts, to have an academic orientation. A school with
a 2-year developmental kindergarten program was excluded from this in-
vestigation. Initial information about the developmental kindergarten
children is provided elsewhere (Mantzicopoulos & Morrison, 1990).

Results
Retained and promoted kindergartners were compared in two ways using
(a) same-age comparisons and (b) same-grade comparisons. For the same-
age comparisons, retained and promoted students were the same age but
the promoted group was a grade level ahead of the retained group. Thus,
children in the two groups took different levels of the academic achieve-
ment tests, and their performance was evaluated on the basis of different
norm groups. For the same-grade comparisons, both retained and promoted
children were in the same grade but the retained children were a year older.
An outline of the study design, which was adopted from Peterson et al.
(1987), is presented in Table 3.
Academic and behavioral outcomes were examined using repeated
measures analyses. A planned model was adopted for these analyses
(Marascuilo & Levin, 1983). Two analyses (same-age and same-grade) were

189

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison

Table 3
Outline of the Study Design8

Group Year 1 Year 2 Year 3 Year 4

Retained Grade K ^ r Grade K ^ r Grade 1 ^~ Grade 2


Promoted Grade K ^ ^ Grade 1 - ^ " " ^ Grade 2 ^^

T h e setup in this table illustrates same-age comparisons within each year of the study.
Diagonal lines indicate same-grade comparisons.

performed for each investigated outcome (Total Reading, Total Math, Con-
duct Problems, Attention Problems-Immaturity, Anxiety Withdrawal, Motor
Excess, Psychotic Behavior). Thus, a total of 14 analyses were made. The
alpha level was split across the 14 analyses so that each F-statistic was
evaluated at alpha = .004 (.05/14). For each analysis, group status (retained
vs. promoted) was the between-subjects variable. Within-subjects variables
were (a) age group during the year in which the test was taken (year 1, year
2, or year 3) or (b) grade at the time of test (second year in grade K, first
grade, or second grade). Significant within-subjects interactions (group status
by year or grade) were examined using post hoc contrasts within each year
or grade.

Academic Effects: Same-Age Comparisons


Results for all academic comparisons are presented in Table 4. The analyses
for same-age comparisons revealed similar trends for both reading and math.
In reading, retained children performed almost a standard deviation
(M = 0.79) above the mean during their second year in school. Post hoc
analyses indicated that this mean was significantly higher (F = 20.60,
p < .0001) than that of their promoted same-age peers (M = 0.14) who were
in first grade, taking a different level of the test during the same time.
However, the advantage of the retained group disappeared as soon as they
went to first grade. At that time, their means were identical to those of the
promoted group (M = 0.22 for both groups). Note that the promoted group
scores were also above average despite their below-average kindergarten
scores at the beginning of the study.
A similar pattern was observed with achievement in mathematics.
Specifically, the retained group outperformed the promoted group during
the second year of school (M = 0.44 for the retained group and M = - 0.08
for the promoted group; F = 11.28, p< .001). Again, as with reading
achievement, the retained group's advantage disappeared during the third
year of school.
190

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
Table 4
Means, Standard Deviations, and F Values of
Retained and Promoted Children on Academic Measures

Retained51 Promoted*
M M F F
Test Grade (SD) (SD) Same-Age Same-Grade

Reading K -0.38 -0.36 ns ns


(0.76) (0-59)
K2 0.79\ -0.36b — 58.78**
( 0 . 7 6 ) ^\ ( 0 . 5 9 )
1 0.22\ ^^ n i A ».9n fcn* *
u . I^T * zu.ou ns
(0.82) v . (0.71)
2 0.16 ^ n ??
u.zz —**ns ns
(0.93) (0.91)
Math K -0.45 -0.35 ns ns
(0-73) (0.64)
K2 0.44\ -0.35 b — 19.06**
(0.81) ^\ (0.64)
, 1 1 OQ *
1 0.13\ — U.Uo ^ 1 1 .Zo ns
( 0 . 7 7 ) ^\ (0.78)
2 0.20 ^ \ f ) ?3
(1.00) (0.92) —**ns ns

Note. Same-grade analyses compare the means across each grade level. The means used for
the same-age comparisons, past thefirstyear in kindergarten, are connected with straight lines.
W = 53. bPromoted children had only one year in kindergarten. Their mean at kindergarten
is placed here as it was used for same-grade comparisons.
*p< .001; **p< .0001

Academic Effects: Same-Grade Comparisons


The analyses of same-grade comparisons revealed a pattern similar to that
observed in same-age comparisons. On both reading and mathematics,
retained children differed significantly from their promoted peers during
their second year in kindergarten (F = 58.78 and F = 19.06, p < .0001 for
reading and math, respectively). This advantage of the retained group did
not hold during first or second grade. Both groups had similar achievement
patterns. On both reading and math, their means were close to the national
mean at the end of first and second grade.
It should be noted that in Marin County mean achievement test scores
are above the national average. However, the academic performance of both
promoted and retained children in this study was below average within
their districts. Support for this statement can be obtained from an examina-
tion of the sample's standardized (z) reading and math achievement test
191

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
scores using local norms. These data are a standardized conversion of
Table 4 and are presented in Table 5. Locally, both groups of children are
considered to be at-risk despite the fact that they are making normal prog-
ress when compared to national norms. It is evident that at kindergarten
both groups are approximately one standard deviation below the mean on
reading (z = -0.76 and z = -0.96 for the promoted and retained groups,
respectively). This below-average performance continues until the end of
second grade and remains approximately half a standard deviation below
the local mean. Similar patterns are seen in math performance where again
both groups are lagging approximately 1/2 standard deviation below their
Marin County classmates.

Investigation of Behavioral Effects


Same-age and same-grade analyses were conducted on the five subscales
of the RBPC. Results are presented in Table 6. Only the comparison on
the Attention Problems-Immaturity subscale reached significance (F = 9.70,
p< 0.001). This result suggested that the retained students (M = 1.33,
SD = 1.56) were rated by their teachers as demonstrating significantly more
immature behaviors and attention problems than their same-age peers
(M = 0.32, SD = 1.04) during their first year in kindergarten. This initial
difference faded during the second kindergarten year. Subsequently, the
two groups were not rated differently by their teachers as indicated by both
same-age and same-grade contrasts.
Since the groups were unequal on the incidence of inattention in kin-
dergarten, the significant drop in attention problems manifested by the re-
tained group later on was difficult to interpret. An attempt was, therefore,
made to select another promoted sample that would be matched to the

Table 5
Means and Standard Deviations of Retained and Promoted Children
on Academic Measures Using Local Norms

Retained2 Promoted2
Grade M SD M SD

K -0.96 0.76 -0.77 0.58


K2 0.07 0.91
1 -0.43 1.15 -0.61 0.90
2 -0.59 1.05 -0.51 0.99
K -0.48 1.65 -0.37 1.55
K2 -0.16 0.86
1 -0.28 0.97 -0.50 0.98
2 -0.65 0.99 -0.59 0.91

W = 53

192

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention

Table 6
Means, Standard Deviations, and F Values of
Retained and Promoted Children on RBPC Subscales

Retained2 Promoted b
>BC p p
ibscale Grade M SD M SD Same-Age Same-Grade

induct K 0.63 1.67 0.79 1.30 ns ns


disorder: K2 0.22-^1.26 0.79 c 1.30c — ns
1 - 0 . 0 1 ^ 1 . 0 6 -- 0 . 1 3 _ 1 . 1 7 .ns ns
2 - 0 . 3 2 0.67-— 0.26—0.89 .ns ns
tention K 1.33 1.56 0.32 1.04 9.70* 9.70*
problems- K2 -0.1W0.94 0.32 c 1.04c — ns
immaturity: 1 -0.19^1.05^^ 0.26_1.01 ^ns ns
2 - 0 . 1 4 0 . 9 5 -- ^ 0 . 0 2 — 1 . 0 9 .ns ns
lxiety K 1.19 2.17 0.45 1.36 ns ns
withdrawal: K2 0.21 - 4 . 4 4 0.45 c 1.36c — ns
1 - 0 . 1 7 ^ 1 . 0 4 -- 0 . 5 1 _ 0 . 6 0 ^ns ns
2 - 0 . 2 8 0 . 7 1 -— 0 . 3 2 _ 0 . 9 4 .ns ns
otor K 0.74 1.42 0.20 1.51 ns ns
excess: K2 0.14^1.19 0.20 c 1.51 — ns
1 - 0 . 0 3 ^ 1 . 1 2 -^ - 0 . 3 1 — 0 . 7 5 .ns ns
2 - 0 . 3 7 1.24-— 0.45-^1.10 .ns ns
;ychotic K 0.38 1.73 - 0.02 0.80 ns ns
behavior: K2 - 0 . 0 2 ^ 1 . 0 5 - 0 . 0 2 c 0.80 c — ns
1 0.31-.1.77- ^ - 0 . 1 5 — 1 . 1 3 .ns ns
2 - 0 . 2 4 0 . 5 8 -^ - 0 . 1 1 _ 1 . 1 2 .ns ns

7te. Same-age comparisons, past thefirstyear of kindergarten, are indicated by straight lines.
r
= 32. bN = 30. cPromoted children only had one year in kindergarten. Their mean at
ndergarten is placed here as it was used for same-grade comparisons.
i < .001

tained group on attention problems also. This attempt was unsuccessful.


was, thus, impossible to select a promoted sample with extreme ratings
i attention problems that would simultaneously retain its similarity to the
tained group on all the matching variables of the study.
Given these circumstances, it was important to conduct an examina-
:>n of the distributions of the two groups on attention problems. A graph
:
these distributions is presented in Figure 1.
Inspection of Figure 1 suggests that the two distributions are rather
tnilar in shape. However, the distribution of the attention scores for the
tained group is more extreme, as indicated by differences in the two tails.
further examination of the distributions of scores obtained by males and
males in each group showed that males were not overrepresented at the

193

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison

| Promoted

S
I I Retained

Attention Problems (z-scores)


Figure 1. Distribution of attention problems for retained and promoted
students at the end of kindergarten

more deviant end of the continuum. Scores for both males and females were
distributed across the range of scores for each distribution. It thus appears
that attention problems may have been an important factor in retention
decisions for a subgroup of children. It may also be the case that teacher
ratings and retention decisions were influenced by teacher views on what
constitutes normal and abnormal behavior.

Discussion
Previous studies exploring the longitudinal effects of elementary school
retention demonstrated that grade repetition is not beneficial for at-risk
students (Holmes, 1989; Holmes & Matthews, 1984; Jackson, 1975). These
studies indicated that the practice is often linked to negative academic, social,
and emotional outcomes in both the lower and upper grades (Holmes, 1989).
However, proponents of early retention have argued that retention is
beneficial when it is applied before the child experiences failure in elemen-
tary school.
The results of the present study failed to support this argument.
Children retained at kindergarten did not demonstrate any lasting academic
gains in reading or mathematics. Only during their second year in kinder-
garten did retained students outperform their promoted counterparts. Since
this advantage was not sustained, one can conclude that retained children
obtained inflated achievement scores the second time around because they

194

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
(a) were tested with the same test at the end of their second year of
kindergarten and/or (b) were a year older at the time of the test.
In general, the results of this study are in agreement with those ob-
tained by Banerji (1990) in her investigation of the longitudinal effects of
a 2-year developmental kindergarten. Banerji's subjects in the developmental
kindergarten program maintained an academic advantage over their pro-
moted peers until the end of first grade. This advantage was lost by the
end of second grade. In contrast, retained children in the present study
failed to maintain their academic achievement gains beyond their second
year in kindergarten. As in the study by Banerji (1990), children in this study
did not lag behind their promoted peers in either same-age or same-grade
comparisons.
Although the academic effects of kindergarten retention were fairly
clear, the behavioral outcomes were much more difficult to interpret. A
constraint was that the availability of the behavioral data depended on each
student's risk status, as defined by SEARCH. The findings, therefore, may
not generalize to the full range of kindergarten children recommended for
retention. At the same time, the comparison of the subsample to the original
sample did not indicate any substantial deviations on the matching variables.
This information suggests that the overall nature of the subsample was not
affected significantly.
In this subsample, retained and promoted children were not signifi-
cantly different in the incidence of conduct problems, hyperactivity, and
anxious, withdrawn, or psychotic behaviors. However, during their first
year of kindergarten, retained children were rated by their teachers as be-
ing significantly more inattentive and immature than their promoted
counterparts. The noted high rates of inattentive, distractible behaviors
sharply declined during the second year of kindergarten. At that point, re-
tained and promoted children received similar ratings, within the normal
range, on both same-age and same-grade comparisons. This result held
through the end of second grade. Could this improvement in behavior be
attributed to retention? Would it be reasonable to assume that an extra year
of kindergarten gave children the chance to mature socially and improve
behaviorally?
Although it may appear that the behavioral results are favoring reten-
tion, the answers to the above questions are far from clear. One must be
reminded that retained and promoted subjects were not matched on
behavior problems. It was impossible to obtain a sample of promoted
children who would be similar to their retained peers on attention problems
as well as on academic, perceptual, gender, and socioeconomic character-
istics. As a consequence, it is not known whether a similar decline in
inattention-immaturity would have occurred if the retained children had
been socially promoted. Simply stated, regression to the mean effects could
not be ruled out.

195

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
Another possible interpretation of the behavioral results may be that
the data simply reflect the operation of teacher bias. It is not unreasonable
to expect that kindergarten teachers who believe in retention for "unready"
children may be more predisposed to rating normal behaviors as highly
immature. If this were the case, the drop in attention problems at first grade
may be a reflection of a change in teacher judgment.
The task of interpreting the behavioral data is made more complex by
the scarcity of retention research with samples of children with pretest data
on all characteristics under longitudinal investigation. It is not uncommon
for studies to examine multiple, long-term outcomes of retention. However,
many such studies do not establish the initial comparability of retained and
promoted children on all measures. An example is found in a well-designed,
recent study on the academic and emotional effects of kindergarten reten-
tion conducted by Shepard and Smith (1989). This study reported no dif-
ferences between promoted and retained children on achievement, social
maturity, attention, and self-concept at the end of first grade (Shepard &
Smith, 1989). In this study, retained kindergarten children were very care-
fully matched to a promoted group on demographic characteristics, domi-
nant language, and developmental readiness. However, adjustment out-
comes and self-concept were examined only at the end of first grade. It
is, therefore, not known how retained and promoted children would have
been rated on these variables in kindergarten prior to being retained.
When the data obtained in the present study are considered past
kindergarten (i.e., at the end of first and second grade), they are similar to
those obtained in other investigations (e.g., Shepard & Smith, 1989). At the
same time, the results clearly illustrate Jackson's (1975) criticism of research
on grade retention; he stated that without random assignment there is "in-
adequate assurance that the pupils [are] initially similar in respect to the ac-
tual conditions which precede grade retention" (p. 619). However, since
random assignment is in most cases impossible, every effort should be made
to provide information on all investigated measures prior to retention.
To conclude, the results of the present study are, in general, consis-
tent with those obtained by other investigators using a similar design (Ban-
erji, 1990; Peterson et al., 1987). Thus, it can be reasonably claimed that
kindergarten retention was not helpful to children in the short interval of
this study. Despite the observed reduction in attention problems, among
retained children, no improvement on academic skills was noted. When
compared to other children in their districts, both retained and promoted
children continued to demonstrate subaverage performance until the end
of second grade.
This evidence does not support the conclusion that kindergarten reten-
tion is beneficial for young at-risk children. Furthermore, the literature on
the association between retention and dropping out of school points to
more harmful effects for children in the long run. It is, therefore, imperative

196

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Kindergarten Retention
that we reexamine alternative programs for kindergarten children at-risk
of school failure. Two excellent reviews by Karweit (1989a, 1989b) report
on the effectiveness of a number of early programs and underscore the need
for c o n t i n u e d research in this area.

References
Banerji, M. (1990, April). Longitudinal effects of a two-year developmental
kindergarten program on academic achievement. Paper presented at the an-
nual meeting of the American Educational Research Association, Boston.
Byrnes, D., & Yamamoto, K. (1986). Views on grade repetition, fournal of Research
and Development in Education, 20(\), 14-20.
California Test of Basic Skills. (1982). Monterey, CA: CTB/McGraw-Hill.
Gardner, E., Madden, R., Rudman, H., Karlsen, B., Merwin, J., Callis, R., & Collins,
C. (1985). Stanford Achievement Test. San Diego, CA: Harcourt Brace
Jovanovich.
Grissom, J.B., & Shepard, L.A. (1989). Repeating and dropping out of school. In
L.A. Shepard & M.L. Smith (Eds.), Flunking grades: Research and policies on
retention (pp. 34-63). Philadelphia: Palmer Press.
Hollingshead, A.H. (1957). Two-factor index of social position. Unpublished
manuscript, Yale University, New Haven, CT.
Holmes, CT. (1989). Grade level retention effects: A meta-analysis of research studies.
In L.A. Shepard & M.L. Smith (Eds.), Flunking grades: Research and policies
on retention (pp. 16-33). Philadelphia: Palmer Press.
Holmes, C.T., & Matthews, K.M. (1984). The effects of nonpromotion on the elemen-
tary and junior high school pupils: A meta-analysis. Review of Educational
Research, 54(2), 225-236.
Karweit, N. (1989a). Effective preschool programs for students at risk. In R.E. Slavin,
N. Karweit, & N.A. Madden (Eds.), Effective programs for students at risk (pp.
75-102). Needham Heights, MA: Allyn and Bacon.
Karweit, N. (1989b). Effective kindergarten programs and practices for students at
risk. In R.E. Slavin, N. Karweit, & N.A. Madden (Eds.), Effective programs for
students at risk (pp. 103-142). Needham Heights, MA: Allyn and Bacon.
Mantzicopoulos, P., & Morrison, D. (1990). Characteristics of at-risk children in tran-
sitional and regular kindergarten programs. Psychology in the Schools, 27,
325-332.
Mantzicopoulos, P., Morrison, D.C., Hinshaw, S.P., & Carte, E.T. (1989). Nonpromo-
tion in kindergarten: The role of cognitive, perceptual, visual-motor, behavioral,
achievement, socioeconomic, and demographic characteristics. American
Educational Research fournal, 26( 1), 107-121.
Marascuilo, L.A., & Levin, J.R. (1983). Multivariate statistics in the social sciences:
A researcher's guide. Monterey, CA: Brooks/Cole.
Morrison, D., & Mantzicopoulos, P. (1990). Predicting reading problems at
kindergarten for children in second grade: SEARCH as a screen. Remedial and
Special Education, 17(4), 29-36.
Peterson, S.E., DeGracie, J.S., & Ayabe, C.R. (1987). A longitudinal study of the ef-
fects of retention/promotion on academic achievement. American Educational
Research fournal, 24(\), 107-118.
Quay, H.C., & Peterson, D.R. (1987). Manual for the Revised Behavior Problem
Checklist. Coral Gables, FL: University of Miami.
Shepard, L.A. (1989). A review of research on kindergarten retention. In L.A. Shepard
& M.L. Smith (Eds.), Flunking grades: Research and policies on retention (pp.

197

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015


Mantzicopoulos and Morrison
64-78). Philadelphia: Palmer Press.
Shepard, L.A., & Smith, M.L. (1989). Academic and emotional effects of kindergarten
retention in one school district. In L.A. Shepard & M.L. Smith (Eds.), Flunking
grades: Research and policies on retention (pp. 79-107). Philadelphia: Palmer
Press.
Shepard, L.A., & Smith, M.L. (1987). Effects of kindergarten retention at the end
of first grade. Psychology in the Schools, 24, 346-357.
Shepard, L.A., & Smith, M.L. (1986). Synthesis of research on school readiness and
kindergarten retention. Educational Leadership, 44, 78-86.
Silver, A.A., & Hagin, R.A. (1981). SEARCH: A scanning instrument for the iden-
tification of potential learning disability (2nd ed.). New York: Walker Educa-
tional Book Corporation.
Silver, A.A., Hagin, R.A., DeVito, E., Kreeger, H., & Scully, E. (1976). A SEARCH
battery for scanning kindergarten children for potential learning disability. Jour-
nal of the American Academy of Child Psychiatry, 15(2), 224-239.

198

Downloaded from http://aerj.aera.net at UNIV OF WESTERN ONTARIO on June 8, 2015

You might also like