Family Relations, 57 (July 2008), 335–350. Blackwell Publishing.

Copyright 2008 by the National Council on Family Relations.

Child Depressive Symptoms, Spanking, and Emotional

Support: Differences Between African American and
European American Youth

C. André Christie-Mizell Erin M. Pryor Elizabeth R.B. Grossman*

Abstract: Using data from the National Longitudinal Survey of Youth—Mother and Child samples, we explored
the relationships among child and adolescent depressive symptoms, spanking, and emotional support offered to
youth. We present cross-sectional and change models for both African Americans and European Americans. Find-
ings showed that regardless of race, spanking is associated with more depressive symptoms in the cross-sectional
analysis but does not appear to maintain this relationship over time. With regard to emotional support, depressive
symptoms for African American youth are inversely related to the emotional support their mothers provide for them
in cross-sectional models, but the benefit does not persist in our change models. For European American children
and adolescents, emotional support is associated with fewer depressive symptoms in the short term and over time.

Key Words: child depressive symptoms, discipline, emotional support, parenting, race/ethnicity, spanking.

A plethora of research indicates that spanking chil-
dren as a measure of correction is positively related
to the frequency and intensity of acting out or exter-
nalizing behavior (Gershoff, 2002; Larzelere, 2000;
Molnar, Buka, Brennan, Holton, & Earls, 2003;
Polaha, Lazelere, Shapiro, & Pettit, 2004; Straus,
1991, 2001). Spanking includes activities such as
striking on the buttocks, slapping a child’s hand, or
pinching a child on the extremities (Gershoff; Straus
& Stewart, 1999). Although spanking includes caus-
ing the child pain for the purpose of correction, this
type of discipline does not involve physical harm
(Gershoff). This form of physical discipline can be
distinguished from physical abuse, which consists of
beatings or other forms of physical force that inflict
bodily injury. In the United States, spanking is viewed
as a normative practice; however, physical abuse is not
(Baumrind, 1996; Turner & Muller, 2004). Although
researchers continue to debate whether increased
behavior problems represent a direct detriment of
spanking or more complex family processes (see, e.g.,
Baumrind vs. Gershoff), the positive association
between harsh disciplinary strategies and behavior
problems, especially those related to child aggression,
is no longer in question (Deater-Deckard, Dodge,
Bates, & Pettit, 1996; Grogan-Kaylor, 2005).
The extent to which spanking is positively related
to mental health or internalizing problems among
youth has been less explored (for an exception, see
Turner & Muller, 2004). Existing research does
indicate that children who routinely experience
harsh physical punishment (including spanking) are
at greater risk for depression as well as a host of
other psychological adjustment problems (Straus,
1994; Turner & Muller). Further, research in this
area is less clear about how the process shaping child
outcomes with regard to spanking may vary by race.
It is clear that African American parents are more
likely than European Americans to spank their
children (Deater-Deckard et al., 1996; Giles-Sims,
Straus, & Sugarman, 1995); however, whether higher
levels of spanking among African Americans result

*C. André Christie-Mizell is an associate professor in the Department of Sociology at Kent State University, 330 Merrill Hall, P.O. Box 5190, Kent, OH 44242-0001
(achrist7@kent.edu), and is a licensed family and child psychologist at Efficacy Psychological and Educational Consultants, LLC, Akron, OH 44313. Erin M. Pryor is
a doctoral candidate in the Department of Sociology at the University of Akron, Olin Hall 247, Akron, OH 44325-1905 (emp14@uakron.edu). Elizabeth R.B. Grossman
is a doctoral candidate in the Department of Sociology at the University of Akron, Olin Hall 247, Akron, OH 44325-1905 (eg12@uakron.edu).
336 Family Relations
Volume 57, Number 3
July 2008
in a greater incidence of problematic behavior
remains equivocal (Polaha et al., 2004; Turner &
Muller). With regard to acting out behavior, some
racial comparisons have indicated that European
American children who are spanked exhibit more
behavior problems than their African American
counterparts (Deater-Deckard et al.), whereas other
research has found no race differences (McLoyd &
Smith, 2002). Therefore, a major goal of this
research was to understand whether the relationship
between spanking and depressive symptoms differs
for African Americans and European Americans.
Our investigation contributes to the research on
spanking and child psychological adjustment in
three important ways. First, accounting for relevant
child and family characteristics (e.g., maternal
depressed mood, family structure, socioeconomic
status, age, and gender), we examined the relation-
ship between spanking and depressive symptoms for
youth born to a nationally representative sample of
African American and European American mothers.
Second, we explored the association between the
maternal emotional support provided to the child
and depressive symptoms and whether this emo-
tional support moderated the relationship between
spanking and depressive symptoms. Third, we devel-
oped cross-sectional and change models to establish
the potential immediate and long-term consequen-
ces of spanking and emotional support on child and
adolescent depressive symptoms.
Theoretical Framework
The Stress Process Model
The Stress Process Model (Pearlin, 1999; Pearlin,
Lieberman, Menaghan, & Mullan, 1981) guided
this research. This paradigm includes three essential
parts: stressors, moderators, and outcomes (Pearlin;
Pearlin & McCall, 1990). Stressors refer to events,
difficulties, or problems that inhibit the individual’s
ability to adapt. Moderators are the psychosocial
resources (e.g., emotional support) that either the
individual has present in the environment or is able
to garner to help reduce or eliminate stressors. Out-
comes are the observed effects of stressors after
accounting for the moderating process.
Depression and youth. In this study, the outcome
of interest was child and adolescent depressive symp-
toms. Depression and depressive symptomatology
are the most prevalent mental health problems
for children and adolescents in the United States
(U.S. Department of Health and Human Services,
1999). For instance, prevalence estimates indicate
that between 2 and 4 million children and adoles-
cents between the ages of 9 and 17 suffer from one
or multiple symptoms of depression, with approxi-
mately 2 million suffering from diagnosable major
depression (Moldenhauer, 2006; Shaffer et al.,
1996). Similar to adults, depressive symptoms in
children and adolescents include frequent sadness,
feelings of hopelessness, decreased interest in activi-
ties, persistent boredom, social isolation, low self-
esteem, poor concentration, major changes in eating
and sleeping habits, and suicidal ideation (U.S.
Department of Health and Human Services;
Hoyert, Kochanek, & Murphy, 1999). Youth with a
depressed mood are more likely as adults to be clini-
cally depressed and to experience poor physical
health, difficulties in sexual relationships, suicide,
involvement in criminal behavior, and low socioeco-
nomic attainment (Klein, Lewinsohn, & Seeley,
1997; Nolen-Hoeksema, Morrow, & Fredrickson,
1993; Weissman et al., 1999).
Spanking as a stressor. The onset or presence of
stress may be a result of specific events such as
parental divorce, the sudden job layoff of the fam-
ily’s breadwinner, or loss of important friendships.
Conversely, another type of stressor is the result of
persistent or chronic events such as daily parental
fights, chronic poverty, or the constant use of harsh
disciplinary tactics (Pearlin, 1999). In this study, we
conceptualized spanking as a potentially chronic
stressor—representing a persistent element of the
child’s environment—that may harm child and
adolescent mental health.
According to Straus (2000), parents use physical
discipline such as spanking ‘‘ . . . with the intention
of causing a child to experience pain, but not injury,
for the purpose of correction or control of the child’s
behavior’’ (p. 1110). The potential of spanking to
become a stressor for youth is related to its preva-
lence, which is quite high among American families.
Early studies showed that the overall rate of ever
experiencing spanking is as high as 99% (Sears,
Maccoby, & Levin, 1957), and more contemporary
studies indicate rates between 90 and 95% (Giles-
Sims et al., 1995; Straus & Stewart, 1999). Younger
children appear to bear the brunt of physical disci-
pline in the home, and spanking appears to peak in
the toddler years, with the vast majority (.90%)
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
experiencing physical discipline. The rate of spank-
ing then declines steadily after age 5 from approxi-
mately 50% to a low of roughly 10% by age 17
(Straus & Stewart). Furthermore, although parents
from a variety of socioeconomic, cultural, and ethnic
backgrounds employ spanking, research indicates
that children who are young (especially in the tod-
dler years), male, southern, economically disadvan-
taged, inner-city residents, or have psychologically
distressed (e.g., high levels of depressive symptoms)
parents tend to experience this type of physical pun-
ishment with greater frequency (Larzelere & Kuhn,
2005; McLoyd, 1990; Straus & Stewart).
The constant use of spanking may compromise
a child’s mental health and connect to depressed
mood for at least two reasons. First, parents might
use spanking for the acting out behavior or the sul-
len attitude that is associated with child and adoles-
cent depressive symptoms. When such behaviors are
a part of the experience of depressed mood and are
not under the control of the child, parental strategies
such as spanking may further stress the child’s adap-
tive capacity and worsen behaviors. After all, spank-
ing is the infliction of pain as a method of correcting
behavior, and youth who are acting out as a result of
depressive symptoms are trying to adapt to psycho-
logical pain. Therefore, parental strategies, such as
spanking, have the potential to cumulatively add to
and exacerbate existing depressive symptoms. Sec-
ond, effective treatment modalities for depression
and depressive symptoms emphasize the need to
demonstrate compassion as a method of encouraging
the individual’s sense of control and purpose that
have been compromised. Therefore, harsh tactics
such as spanking are stressful and antithetical to the
emotional support that the child needs to combat
depressive symptomatology.
Emotional support as a psychosocial resource. In
the Stress Process Model, psychosocial resources
emanate from personal characteristics (e.g., positive
self-concept) or interpersonal relationships (e.g.,
maternal emotional support) that buffer the individ-
ual from stress. Psychosocial resources have the abil-
ity to either reduce or eliminate stressors, thus
improving mental health. In this study, we focused
on whether the emotional support provided by the
child’s mother reduced the ill effects of spanking on
the psychological well-being of youth. Emotional
support includes physical affection, affirming atten-
tion, and the communication of encouragement.
Research indicates that high levels of emotional
337
support or parental warmth is positively related to
many child and adolescent outcomes, including high
academic achievement, lower rates of behavior prob-
lems, and more positive self-concept (Christie-
Mizell, 2004; McCord, 1997; McLoyd & Smith,
2002; Smith & Brooks-Gunn, 1997). Specifically,
the emotional support provided to children moder-
ates behavior problems by reducing the deleterious
effects of spanking (McLoyd & Smith). The theoret-
ical reasoning behind this pattern is that emotional
support provides the context in which physical disci-
pline is administered. In the context of high parental
warmth, the harshness of spanking is minimized and
positive outcomes are more likely to occur (Deater-
Deckard & Dodge, 1997; DeVet, 1997; Kelley,
Power, & Wimbush, 1992; Rohner, Bourque, &
Elordi, 1996). In the current study, we addressed
the question of whether emotional support moder-
ates the influence that spanking has on youth
depressive symptomotology.
Race Differences and the Implications of Spanking
Race/ethnicity and culture may play a significant
role in how the use of spanking and the provision of
emotional support translate into child and adoles-
cent well-being. African Americans are more likely
compared to their European American counterparts
to employ spanking as a disciplinary strategy
(Deater-Deckard et al., 1996; Gershoff, 2002).
McLoyd (1990) developed a model that indicates
that harsh strategies, such as spanking, may be more
prevalent among racial minorities because of the
strains and challenges inherent in their daily lives.
For example, African American mothers are more
likely to be single parents, with low income, unem-
ployed, educationally challenged, and to have larger
families, compared to Whites (McLoyd). The frus-
tration and high levels of stress associated with par-
enting in this context may simply be a disciplinary
strategy directed at quick correction through the use
of spanking.
With regard to race and depressive symptoms,
one firmly established finding is that African Ameri-
cans do disproportionately face more of the stressors
related to depressive symptomatology, including but
certainly not limited to chronic poverty, unemploy-
ment, violence, and discrimination (Christie-Mizell,
Steelman, & Stewart, 2003; Kessler et al., 1994;
King & Williams, 1995). Therefore, African Ameri-
cans are more likely to experience more depressive
338 Family Relations
Volume 57, Number 3
July 2008
symptoms but not necessarily clinical levels of
depression. Compared to White youth, one specific
risk factor for African American children and adoles-
cents for depressive symptoms may be the provision
of maternal emotional support. Although emotional
support as a psychosocial resource is manifest in the
relationship between mother and child, research
indicates that it may be more difficult for some
mothers to provide a context that involves high
levels of parental warmth. For example, McLoyd
and Smith (2002) found race-based differences in
the level of emotional support offered to children.
This pattern is partially explained by the fact that
the provision of emotional support is predicated
upon socioeconomic factors such as education and
income and also the fact that racial minorities are
more likely to be disadvantaged (Mott, 2004).
Mothers with greater resources have the education,
time, and social support needed to build home envi-
ronments in which their children benefit from
higher levels of warmth and support (Caldwell &
Bradley, 1984; McLoyd, 1990; Mott, 2004; Parcel
& Menaghan, 1993).
Summary and Hypotheses
Using the Stress Process Model as a guide, this study
investigated the influence of spanking as a stressor
that may connect with depressive symptoms among
children and adolescents. Consistent with the theo-
retical assumptions of our conceptual model, we also
explored whether the emotional support offered in
the home moderates the deleterious effects of spank-
ing on child and adolescent mental health. This
study adds to the growing body of research literature
that seeks to understand the complex process
through which parental behaviors affect the well-
being of children.
To assess the relationship between spanking and
child well-being, we estimated both cross-sectional
and change models. This study had three main
hypotheses. First, we anticipated that spanking was
positively related to child and adolescent depressive
symptoms. Second, we hypothesized that maternal
emotional support decreases symptoms of depres-
sion among children. Third, we expected that emo-
tional support would moderate the relationship
between spanking and depressive symptoms. That
is, we were hypothesizing that maternal emotional
support would decrease the otherwise deleterious
association between spanking and children’s mental
health. In addition to these main hypotheses,
another important task of this research was to
develop models that enabled us to assess whether the
cross-sectional and change patterns among child
depressive symptoms, spanking, and emotional sup-
port vary by race.
Method
Data Set
We used data from the National Longitudinal Sur-
vey of Youth (NLSY) and National Longitudinal
Survey of Youth Child (NLSY-C) sample. The data
in the NLSY were collected under the auspices of
the U.S. Departments of Labor and Defense (Center
for Human Resource Research, 2002). The original
NLSY sample overrepresented racial minorities and
economically disadvantaged White youth. Respon-
dents were interviewed annually from 1979 to 1994
and biennially after 1994. Initial ages ranged from
14 to 22 years.
The NLSY-C is a survey of children born to the
women of the NLSY. These children have been
interviewed biennially since 1986. In each year of
the NLSY-C, mothers from the NLSY, and their
children, were interviewed. Assessments were made
of the child’s cognitive ability and social develop-
ment as well as externalizing and internalizing
behavior problems. Additionally, the NLSY provides
information on the mother-child relationship and
the quality of the home environment. With the
exception of a Time 1 (T1) measure of child and
adolescent depressive symptoms from the 1992
wave, all variables were derived from the 1994 wave
of the NLSY and NLSY-C. We were restricted to
these years because 1992 and 1994 were the only
years that maternal depressive symptoms, one of our
control variables, were assessed in the NLSY. In
1994, the mothers in the sample were 29 – 37 years
old.
We used several techniques to explore the 452
cases that had missing data on one or more of the
study variables. The results of ‘‘complete cases’’ anal-
yses did not differ from our auxiliary analyses (e.g.,
selection models and multiple imputation). There-
fore, the complete cases analyses presented below
have weighted data to correct for the oversampling
of economically disadvantaged youth. The total
sample size (N) for the study includes 713 African
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
American and 1,139 European American children
and adolescents ranging in age from 6 to 14 years.
Table 1 presents the means, standard deviations,
and correlations for all variables used in this study.
Measures
Dependent variable: Depressive symptoms. Child
and adolescent depressive symptoms was the depen-
dent variable for this study. We measured this vari-
able in two waves, 1992 (T1) and 1994 (Time 2
[T2]), of the NLSY-C. In these data, child depres-
sive symptoms were a 5-item subscale of the Behav-
ior Problems Index (BPI; Peterson & Zill, 1986).
The BPI is a 28-item scale designed to assess typical
childhood behavior syndromes, rather than unusual
behavior that may indicate serious pathology. Sev-
eral of the BPI items were derived from the Achen-
bach Behavior Checklist (Achenbach & Edelbrock,
1981, 1983) and other child behavior scales (Graham
& Rutter, 1968; Kellam, Branch, Agrawal, &
Ensminger, 1975; Rutter, 1970).
Relying on maternal reports, the five items from
the BPI used to measure depressive symptoms asked
whether the child (a) ‘‘experiences sudden changes
in mood/feelings’’; (b) ‘‘feels/complains that no one
loves him/her’’; (c) ‘‘is too fearful or anxious’’; (d)
‘‘feels inferior or worthless’’; or (e) ‘‘is unhappy, sad,
or depressed.’’ Each item ranges from 1 (not true) to
3 (often true). In multiple studies, NLSY researchers
have confirmed the reliability (a . .70) and factor
analytic strength of this depressive symptoms sub-
scale (Center for Human Resource Research, 2002;
Mott, 1998). Several versions of the depression scale
exist within the NLSY-C data, including raw scores,
standardized scores, and percentile ranks. Similar
to other studies using these data (e.g., Parcel &
Menaghan, 1994), we used age-specific ‘‘standard’’
scores that were normalized to a mean of 0 and a
standard deviation of 1. Our T1 and T2 measures
were correlated at .466 (p , .001) and .580 (p ,
.001) for African Americans and European Ameri-
cans, respectively (Table 1).
Independent variables: Spanking and emotional
support. Our two main independent variables were
spanking and the level of emotional support that
each mother provided for her child. The spanking
variable asked each mother to report how many
times she spanked her child in the past week. For
the entire sample, spanking ranged from 0 to 30,
with an average of 0.85 for African Americans and
339
0.34 for European Americans. These scores differed
significantly (p , .001) and were positively corre-
lated with depressive symptoms for both groups.
Maternal emotional support for the child was a 5-
item subscale of the Home Observation for Mea-
surement of the Environment (HOME) Scale,
developed by Bradley and Caldwell (1980) and
Bradley, Caldwell, and Elardo (1977). The HOME
observation relied on data reported by each mother
and observations made by the interviewer during
the interview visit. The five items that measured
emotional support included four observations by the
interviewer that judged how pleasantly the mother
responded to the child during the interview, whether
the child was encouraged to talk, whether the
mother responded positively (e.g., use of a calm
voice) to the child even during misbehavior, and
whether the mother treated the child with physical
affection (e.g., caress, kiss, hug) during the visit. Fur-
ther, information was collected from each mother
on the frequency of time that she spends with her
child (e.g., meals together). The interviewers’ obser-
vations were adjusted for the frequency of time
together. The scale was internally standardized by
the NLSY-C research team and was coded for this
study to range from 1 (lower emotional support) to
10 (higher emotional support; see Center for Human
Resource Research, 2002, for a detailed description
of the HOME Scale and the coding scheme). Afri-
can American mothers (M ¼ 3.69) in our sample
offered significantly less (p , .001) emotional sup-
port to their children, compared to their European
American counterparts (5.56). However, for both
African Americans (r ¼ 2.12, p , .01) and European
Americans (r ¼ 2.17, p , .001), emotional support
was inversely related to child and adolescent depres-
sive symptoms.
Control variables. In all models estimated below,
we controlled for child’s age (years), which ranged
from 6 to 14 years, with a mean of 10.21 for African
Americans and 9.83 for European Americans. Forty-
nine percent of the African American sample and
50% of the European American sample were male.
In terms of family structure, we compared single,
never married mothers to all other family forms. For
our samples, 31% of African American mothers
were never married compared to only 3% of Euro-
pean American mothers. Further, we controlled for
the number of dependent children in the home. Sig-
nificantly (p , .001) varying by race, the average
number of dependent children was 2.81 in the
 340

Table 1. Descriptives and Intercorrelations Among Study Variables

Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14
African Americans
1. Child depressive
symptoms—T2
(BPI subscale)
2. Child depressive .47***
symptoms—T1
(BPI subscale)
3. Child’s age (years) .04 .07
4. Male (1 ¼ yes) .03 .02 .04
5. Mother-headed home .06 .05 .04 .06
(1 ¼ never married)
6. Family size (count) .00 .00 2.02 .03 .02
7. Central city residence 2.06 .02 2.01 .00 .07 .02
(1 ¼ yes)
8. Region (1 ¼ south) .01 2.06 2.02 .04 2.14*** 2.11** 2.21***
9. Household income 2.13*** 2.11*** 2.08* 2.02 2.31*** 2.09* 2.06 2.04
(logged)
10. Maternal employment 2.09* 2.10* 2.03 .01 2.24*** 2.28*** 2.04 .07 .36***
(1 ¼ yes)
11. Maternal education 2.21** 2.09* 2.10** 2.06 2.18*** 2.25*** 2.02 2.04 .37*** .34***
(years)
Family Relations
Volume 57, Number 3
July 2008

12. Maternal depressive .18*** .21*** .01 2.02 .04 .06 2.06 2.04 2.21*** 2.14*** 2.15***
symptoms (7-item
CES-D)
13. Frequency of spanking .12** .11** 2.13*** .13*** .11** .09* .13*** 2.01 2.07* 2.06 2.09* .05
14. Maternal emotional 2.12** 2.11** .06 2.02 2.22*** .02 2.05 .02 .21*** .06 .09* 2.09* 2.12**
support for child
(HOME subscale)
M 0.00 0.00 10.21 0.49 0.31a 2.81a 0.22a 0.61a 2.84a 0.59a 12.33a 5.53a 0.85a 3.69a
SD 1.00 1.00 2.41 0.50 0.46 1.18 0.42 0.49 0.95 0.49 1.79 4.57 2.35 2.90
European Americans
1. Child depressive
symptoms—T2
(BPI subscale)
2. Child depressive .58***
symptoms—T1
(BPI subscale)
3. Child’s age (years) .16*** .15***
4. Male (1 ¼ yes) .00 2.05 2.01
5. Mother-headed home .05 .06 .02 2.02
(1 ¼ never married)
6. Family size (count) 2.03 2.06 2.01 2.02 2.09
7. Central city residence .07* .03 .02 .04 2.01 .01
(1 ¼ yes)
8. Region (1 ¼ south) 2.04 2.04 .06* .02 2.08* 2.10*** .07*
9. Household income 2.15*** 2.14*** 2.08** .00 2.19*** .05 2.06 2.09**
(logged)
10. Maternal employment .01 2.01 .09** .00 2.06 2.14 .00 .00 .13***
(1 ¼ yes)
11. Maternal education 2.06* 2.05 2.20*** .00 2.05 .05 2.06 2.03 .32*** .00
(years)
12. Maternal depressive .23*** .28*** .08** .02 .10*** 2.09** .01 .00 2.24*** 2.07* 2.20***
symptoms (7-item
CES-D)
13. Frequency of spanking .09** .08** 2.16*** .06* 2.03 .02 2.03 .04 2.11*** 2.06 2.06 .08**
Child Depressive Symptoms and Spanking
Christie-Mizell et al.

14. Maternal emotional 2.17*** 2.22*** 2.03 2.04 2.08** .06* 2.02 2.01 .26*** 2.01 .12*** 2.22*** 2.13***
support for child
(HOME subscale)
M 0.00 .00 9.83 0.50 0.03a 2.51a 0.04a 0.29a 3.51a 0.70a 12.79a 4.14a 0.34a 5.56a
SD 1.00 1.00 2.40 0.50 0.17 1.03 0.20 0.45 0.81 0.46 1.86 4.24 0.96 2.68

Note. The models are for a National Longitudinal Survey of Youth Child sample of African American (N ¼ 713) and European American (N ¼ 1,139) children and adolescents. BPI ¼ Behavior Problems Index; CES-D ¼ Center for the
Epidemiologic Study of Depression Scale; HOME ¼ Home Observation for Measurement of the Environment Scale; T1 ¼ Time 1; T2 ¼ Time 2.
a
Significantly differs between Africans Americans and European Americans at p , .01.
*p , .05. **p , .01. ***p , .001.
341
342 Family Relations
Volume 57, Number 3
July 2008
African American families and 2.51 in the European
American families. Child depressive symptoms were
inversely related to the number of children in the
home for European Americans. We also accounted
for place of residence and region. In terms of resi-
dence, we created dummy variables to distinguish
central/inner city (i.e., urban core), suburban, and
rural locations. We compared those who lived in the
central city versus all other residences. In terms of
region, we contrasted those who reside in the south-
ern region to all other regions (e.g., Northeast, Mid-
west). In our sample, significantly more African
Americans inhabited the central city (22 vs. 4%) and
the southern region (61 vs. 29%), compared to their
White counterparts. Our preliminary sensitivity
analyses as well as the empirical and theoretical work
of others (see, e.g., Straus & Stewart, 1999, or
McLoyd, 1990) justified these decisions regarding
residence and region.
We further controlled for household income. For
African Americans, the mean household income was
$24,448, whereas it was $44,511 for European
Americans. For the analyses, household income was
logged to correct for skewness and to prompt homo-
geneity of error variance (Christie-Mizell, 2006).
This log transformation resulted in values of 2.84
and 3.51 for African Americans and European
Americans, respectively. These values differed signif-
icantly by race and were also significantly related to
child and adolescent depressive symptoms for both
groups.
The models developed for this project also held
constant other socioeconomic factors including years
of maternal education and employment. African
American mothers reported significantly (p , .001)
lower education (12.33 vs. 12.80 years) and rates of
employment (59 vs. 70%) compared to European
American mothers. For the African American
mothers in our sample, both the years of education
(r ¼ 2.11, p , .01) and employment (r ¼ 2.08,
p , .05) were inversely correlated with our main
dependent variable—child depressive symptoms
(T2). Neither of these variables appeared to be sig-
nificantly associated with depressive symptoms
among European American youth.
Finally, we accounted for maternal depressive
symptoms as measured by a 7-item version of the
Center for the Epidemiologic Study of Depression
Scale (Radloff, 1977). The well-known and vali-
dated scale estimates symptoms of depression in
community populations but does not indicate
a diagnosis of clinical depression. However, the mea-
sure does distinguish between distressed and nondis-
tressed individuals and has been highly correlated
with other depression scales (Weissman, Scholomskas,
Pottenger, Prusoff, & Locke, 1977). The seven
items included in each of our scales were all prefaced
by the interviewer with, ‘‘During the last week . . . ’’
and included the following: ‘‘I had a poor appetite,’’
‘‘I had trouble keeping my mind on tasks,’’ ‘‘I felt
depressed,’’ ‘‘Most everything took extra effort,’’ ‘‘I
felt sad,’’ and ‘‘I could not get going.’’ We scored
each item of the scale ranging from 0 (low) to 3
(high). These seven items were summed, yielding
a scale ranging from 0 (lower depression) to 21
(higher depression). The significantly different (p ,
.001) mean scores was 5.53 for African American
mothers and 4.14 for European American mothers.
These depression scales exhibited high reliability,
with Cronbach’s alphas of .81 and .84 for African
American and European American mothers, respec-
tively. Maternal depressive symptoms were positively
associated with child depression for both African
Americans (r ¼ .21, p , .001) and European Amer-
icans (r ¼ .28, p , .001).
Analytic Strategy
We estimated both cross-sectional and change mod-
els to evaluate the associations among spanking,
emotional support, and child depressive symptoms.
One of our principle aims was to explore race differ-
ences in the process shaping child and adolescent
depressive symptoms. Therefore, instead of estimat-
ing multiple interactions, which tend to complicate
interpretation, we estimated each model separately
for African Americans and European Americans.
To test for difference in the size of associations
across groups when coefficients are significant for
both groups, we employed a comparison t statistic
(Christie-Mizell, 2006). The equation for the com-
parison t statistic can be written as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
bðx1 Þ2bðx2 Þ= ½SE ðx1 Þ2 1 ½SE ðx2 Þ2 ;
where b(x 1) represents the unstandardized coeffi-
cient for variable x in the first equation and b(x 2) is
the coefficient for variable x in the second equation.
Further, SE(x 1) is the standard error for variable x in
the first equation, whereas SE(x 2) is the standard
error for variable x in the second equation.
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
The cross-sectional analysis included three mod-
els. The first model included all the control varia-
bles, which prior research has demonstrated to be
related to both our outcome measure (i.e., child
depressive symptoms) and our main independent
variables (i.e., spanking and emotional support).
Our second and third models progressively added
the (Model 2) frequency of spanking and (Model 3)
emotional support for the child. The second and
third models allowed us to assess the association
between child depressive symptoms and spanking
with and without the emotional support for the
child in the equation. In these progressive adjust-
ment models, we also assessed and reported the
change in the size of association of significant varia-
bles as we added variables to the model (Stolzenberg,
1980). To test our third hypotheses regarding
whether emotional support moderates spanking, we
estimated interactions (i.e., Spanking  Emotional
Support). The cross-sectional models in this study
took the form
childdepi ¼ b0 1 b1 spank i 1 b2 emosupi
1 b3 CFCi 1 b4 SESi 1 ei ;
where child depressive symptoms (childdepi ) is
a function of frequency of spanking (spanki ), emo-
tional support (emosupi ), controlling for child and
family (CFCi ) characteristics, and socioeconomic
status (SESi ). Scalar variables are represented by
lower-case variables such as spanki, whereas vector
variables are represented by upper-case variables
such as CFCi , indicating that more than one type of
child and family characteristic is embedded in
CFCi .
We also estimated a change model. The expres-
sion of the cross-sectional calculation can be trans-
formed into a change model simply by adding the
term b6childdepT1i , where childdepT1i represents
a T1 measure of childdepi , to the right side of the
equation. Cross-sectional regression analyses are
often criticized for containing findings that may be
a result of omitted variables or specification error
(Christie-Mizell, 2004; Guo & VanWey, 1999).
The inclusion of a T1 measure added a longitudinal
element to our model and corrected for omitted var-
iables. Therefore, our change model provided a more
stringent test of multivariate associations, compared
to the cross-sectional estimations. The T1 measure
of child depressive symptoms increased confidence
that significant relationships (e.g., between spanking
343
and depressive symptoms) accurately reflected the
association of those variables with the outcome mea-
sure rather than simply capturing the effect of earlier
child depressive symptoms. For the purposes of this
research, cross-sectional models are helpful in under-
standing the initiation or onset of child depressive
symptoms. Conversely, the findings in change mod-
els have implications for factors that are important in
understanding depressive symptomatology over time.
When variables are significant in both cross-sectional
and change models, the indication is that such fac-
tors potentially have immediate and long-term con-
sequences for the behavior under study (Parcel &
Menaghan, 1993).
Results
Table 2, Models 1 – 3, shows cross-sectional models
for child and adolescent depressive symptoms. Mod-
els 1a and 1b show a baseline model, which included
child and family characteristics, socioeconomic sta-
tus, and maternal levels of depressive symptoms. In
these models, maternal depressive symptoms ap-
peared to be the only covariate that was related to
African American children’s depressive symptoms.
Maternal depressive symptoms were also related to
higher levels of depressive symptoms for European
American youth. The size of the association between
mother and child depressive symptoms was larger
(t ¼ 2.23, two-tailed test) for European Americans
compared to African Americans. Further, among
European American children and adolescents, age
was positively related to depressive symptoms,
whereas being male, household income, and mater-
nal education were negatively associated with
depressive symptoms.
In Table 2, Models 2a and 2b, spanking was pos-
itively related to depressive symptoms for both racial
groups, and the size of this association did not vary
across groups. That is, regardless of race, the fre-
quency of spanking was related to increases in child
and adolescent depressive symptoms. Although
slightly reduced in the size of the association for both
African Americans and European Americans, mater-
nal depressive symptoms stayed significant, with the
association remaining significantly larger for Euro-
pean Americans (t ¼ 2.22, two-tailed test). More-
over, for European Americans, child’s age retained
its positive association with the outcome measure,
and being male, household income, and maternal
 344

Table 2. Child and Adolescent Depressive Symptoms Regressed on Selected Variables

African European African European African European African European
Americans, Americans, Americans, Americans, Americans, Americans, Americans, Americans,
Model 1a Model 1b Model 2a Model 2b Model 3a Model 3b Model 4a Model 4b
Independent Variables b (SE) b (SE) b (SE) b (SE) b (SE) b (SE) b (SE) b (SE)
Child’s age 0.30 (0.19) 0.74*** (0.16) 0.37 (0.19) 0.82*** (0.16) 0.40* (0.19) 0.82*** (0.16) 0.32 (0.17) 0.36** (0.14)
Male (1 ¼ yes) 0.51 (0.91) 21.56* (0.76) 0.19 (0.92) 21.70 (0.76) 0.22 (0.91) 21.80* (0.75) 0.03 (0.82) 21.60* (0.64)
Mother-headed home 0.26 (1.07) 0.57 (2.30) 0.10 (1.06) 0.88 (2.30) 20.22 (1.07) 0.50 (2.27) 20.09 (0.97) 0.08 (1.93)
(1 ¼ never married)
Number of dependent 20.43 (0.41) 20.50 (0.38) 20.49 (0.41) 20.53 (0.38) 20.45 (0.41) 20.48 (0.37) 20.15 (0.37) 20.49 (0.32)
children (count)
Central city residence 0.72 (1.07) 2.11 (1.93) 0.35 (1.08) 2.33 (1.92) 0.28 (1.08) 2.36 (1.90) 1.11 (0.97) 20.13 (1.62)
(1 ¼ yes)
Resides in the South 21.25 (0.97) 21.59 (0.85) 21.30 (1.00) 21.69* (0.85) 21.30 (0.96) 21.66* (0.84) 21.17 (0.87) 20.71 (0.71)
(1 ¼ yes)
Household income 20.34 (0.63) 21.38* (0.54) 20.33 (0.63) 21.22* (0.54) 20.08 (0.64) 20.68 (0.54) 0.04 (0.57) 20.03 (0.46)
(logged)
Maternal employment 21.29 (1.06) 0.06 (0.84) 21.25 (1.06) 0.10 (0.84) 21.39 (1.06) 20.15 (0.83) 21.23 (0.95) 20.48 (0.71)
(1 ¼ yes)
Maternal education 20.26 (0.29) 20.45* (0.22) 20.21 (0.29) 20.48* (0.22) 20.19 (0.29) 20.50* (0.22) 0.36 (0.27) 20.31 (0.18)
(years)
Maternal depressive 0.53*** (0.10) 0.83*** (0.09) 0.52*** (0.10) 0.82*** (0.09) 0.51*** (0.10) 0.74*** (0.09) 0.35*** (0.09) 0.44*** (0.08)
Family Relations
Volume 57, Number 3
July 2008

symptoms
Frequency of spanking 0.57** (0.21) 1.18** (0.40) 0.54** (0.21) 1.00* (0.40) 0.29 (0.19) 0.34 (0.34)
Maternal emotional 20.34* (0.16) 20.74*** (0.15) 20.19 (0.15) 20.48*** (0.13)
support for child
Child and adolescent 0.43*** (0.03) 0.51*** (0.02)
depressive symptoms—T1
Adjusted R 2 .05 .10 .06 .11 .06 .13 .24 .37

Note. The regression models are for a National Longitudinal Survey of Youth Child sample of African American (N ¼ 713) and European American (N ¼ 1,139) youth. T1 ¼ Time 1.
*p , .05. **p , .01. ***p , .001.
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
education continued to be negatively associated with
increased depressive symptoms. Finally, in Model
2b, residing in the South for European Americans
was also negatively related to children’s depressive
symptoms.
In Table 2, Models 3a and 3b, the maternal
emotional support offered to the child was added to
the model. For African Americans and European
Americans, emotional support was negatively related
to child depressive symptoms and the size of this
association did not differ significantly. Greater emo-
tional support was associated with fewer depressive
symptoms. However, spanking remained significant
and the size of the unstandardized coefficient did
not differ by race. Interestingly, with emotional sup-
port in the model, the size of the association between
maternal depressive symptoms and child depressive
symptoms was still significant but now had a similar
relationship across race (t ¼ 1.71, two-tailed test).
Additionally, including emotional support revealed
a positive association between age and depressive
symptoms for African American youth. As in prior
models, age continued to be positively related to
depressive symptoms for European Americans and
the size of this association did not vary by race. Fur-
ther, being male, residing in the South, and maternal
education continued to be negatively related to
European American children’s depressive symptoms.
Our final estimation was a change model
(Table 2 Models 4a and 4b). The results showed
that the T1 measure of depressive symptoms posi-
tively and significantly influenced child depressive
symptoms at T2 for African Americans and Euro-
pean Americans. Also, the size of the association
between the T1 and the T2 measures of depressive
symptoms was larger (t ¼ 2.21, two-tailed test) for
European Americans compared to their African
American counterparts. Furthermore, regardless of
race, the frequency of spanking was no longer signif-
icantly related to child and adolescent depressive
symptoms. In other words, over time spanking did
not appear to be related to depressive symptoms for
African American or European American youth. For
African Americans, other than the T1 measure of
depressive symptoms, only maternal depressive
symptoms, reduced in its association by 31% (b ¼
0.51 vs. b ¼ 0.35), was significantly related to child
depressive symptoms. For European American chil-
dren and adolescents, the change model, compared
to the final cross-sectional model (Model 3b), shows
that residing in the South and maternal education
345
were no longer significantly related to depressive
symptoms. However, although the size of their asso-
ciations was reduced, emotional support for the
child and being male were still negatively related to
child depressive symptoms, whereas being older and
maternal depressive symptoms retained their positive
associations with child depressive symptoms at T2.
In addition to the results shown in Table 2, we
also estimated interactions between spanking and
emotional support for both subsamples to test for
moderation. The association between spanking and
depressive symptoms was not moderated by the
emotional support provided to the child either in
cross-sectional or change models. These analyses are
available upon request.
Discussion
Using the Stress Process Model as a theoretical
framework, the purpose of this study was to examine
the relationships among child depressive symptoms,
spanking, and emotional support provided to the
child. Also, in the context of both cross-sectional
and change models, a central aim of this research
was to explore whether the process contributing to
depressive symptomatology varies for African Ameri-
can youth compared to European American youth.
The results provide support for the Stress Process
Model and mixed support for our first hypothesis
that predicted a positive relationship between spank-
ing and depressive symptoms. In the cross-sectional
models, spanking was a stressor that was equally
associated with depressive symptoms for African
Americans and European Americans; however, in
the change models, spanking was not a factor that
was significantly related to depressive symptoms,
and this finding diverges from our Hypothesis 1.
This pattern suggests that spanking may be impli-
cated in the initiation or onset (i.e., the cross-
sectional model) of depressive symptoms but is not
a particularly salient factor over time (i.e., the
change model; see, e.g., Christie-Mizell, 2004, or
Parcel & Menaghan, 1993, for similar explanations
of cross-sectional vs. change models).
Our second hypothesis was also derived from the
Stress Process Model and detailed a negative rela-
tionship between emotional support (a psychosocial
resource or moderator of stress) and child depressive
symptoms. Our findings provide support for this
hypothesis in the cross-sectional models for both
346 Family Relations
Volume 57, Number 3
July 2008
African Americans and European Americans. Lower
levels of depressive symptoms are associated with
more emotional support to the extent that such
warmth can diminish the harshness of physical
discipline and encourage more positive outcomes
(Deater-Deckard et al., 1996). However, this nega-
tive association was only maintained in the change
models for European Americans, suggesting that this
group compared to African Americans is advantaged
in terms of how emotional support may be impli-
cated in the reduction of depressive symptoms over
time. There are at least two explanations for this
finding. First, given that European American
mothers provide higher levels of emotional support,
it may be that in the case of spanking and depressive
symptoms, this elevated emotional support persists
over time. Second, in the current study, as well as
prior research (see, e.g., Christie-Mizell et al., 2003),
the factors that shape and are associated with mental
health differ by culture and race. Therefore, we can-
not rule out that although emotional support is an
important, positive factor in the lives of all children,
it may be differentially associated with depressive
symptoms for African American youth compared to
European American youth.
Our third hypothesis consisted of the expectation
that the emotional support offered to children
would moderate the association between spanking
and depressive symptoms. In the cross-sectional and
change models presented above, spanking was not
moderated by emotional support for either African
Americans or European Americans. This finding is
surprising insofar as other scholars (see, e.g.,
McLoyd & Smith, 2002) using these same data, but
different outcome measures, have found that mater-
nal emotional support buffers or moderates the asso-
ciation between spanking and undesirable child
outcomes. Nevertheless, our models include mater-
nal depressive symptoms, one important variable
missing from other research. Furthermore, studies
that have found that emotional support moderates
the relationship between physical discipline have
largely focused on externalizing problems (e.g., anti-
social behavior), rather than depressive symptom-
atology, an internalizing problem (see, e.g., Deater-
Deckard & Dodge, 1997). These differences in our
model may explain why we did not find the pre-
dicted moderation effects. Despite not finding statis-
tical moderation (i.e., a significant interaction term
between emotional support and spanking; Baron &
Kenny, 1986), it would be misleading to say that
emotional support was not connected to how spank-
ing operated in our analysis. For example, in the
cross-sectional analysis (Table 2, Models 3a and
3b), the addition of emotional support to our model
did diminish the positive association between de-
pressive symptoms and spanking by 5% (b ¼ 0.57
vs. b ¼ 0.54) for African Americans and by 15%
(b ¼ 1.18 vs. b ¼ 1.00) for European Americans.
This pattern suggests that although multiplicative
effects do not exist, as suggested by the Stress Process
Model, emotional support buffers youth from symp-
toms of depression.
Other Race/Ethnic Differences
Other differences between African American and
European American youth also emerged from our
analyses. First, child’s age had a significant, positive
relationship with depressive symptoms for European
Americans in both the cross-sectional and the
change models. Increases in age for White children
were associated with increases in depressive symp-
toms, suggesting the importance of early interven-
tion for this group. Second, European American
girls appear to be at greater risk for depression than
their male counterparts. This pattern among Euro-
pean American youth parallels findings from
research on depressive symptoms in adult popula-
tions (Kessler et al., 1994). No gender pattern was
found among African Americans. Third, family size
was a buffer against depression for Whites but not
for Blacks. That is, the more siblings in the home,
the fewer depressive symptoms among European
American youngsters. This finding is notable insofar
as other research has indicated that family size is
inversely related to some positive outcomes for both
parents and children (Christie-Mizell, 2003; Chris-
tie-Mizell et al., 2003; Mizell & Steelman, 2000).
The typical reasoning is that fewer resources are
available to any one child or family member as fam-
ily size increases (Mizell & Steelman). However, it
may be that with regard to depressive symptoms
the opportunity for support and interaction with
siblings results in a negative association with depres-
sive symptoms.
Finally, in our cross-sectional analysis, residing in
the southern region of the United States was associ-
ated with fewer depressive symptoms for White
youth. Although this finding was not present for the
African Americans in our sample, this finding does
parallel national prevalence estimates for youth aged
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
12 – 17 on depressive symptomatology in which the
southern tier of states have overall lower levels of
depression than other regions of the country, espe-
cially after controlling for socioeconomic status
(Substance Abuse and Mental Health Services
Administration, 2007). This finding is especially
intriguing, given that the frequency of spanking, one
of our main independent variables and a mental
health stressor, is higher in the southern region of
the United States (Flynn, 1994, 1996). This result
led us to test whether the association of spanking
might be moderated in some manner by residing in
the South. The interaction terms used to test this
contention were not significant; this additional anal-
ysis is available upon request. Future studies should
further examine the nexus between region of origin,
spanking, and depressive symptoms to further eluci-
date how and why living in the South can be associ-
ated with higher rates of spanking (a detriment to
mental health) and better mental health for Euro-
pean American children and adolescents.
Implications for Practice and Future Research
Findings in this research have three implications
with respect to spanking and child and adolescent
outcomes. First, this work informs the ongoing
debate among researchers, child advocates, and par-
ents about whether spanking is an appropriate form
of correction (see, e.g., Baumrind, 1996). Because of
the cross-sectional relationship between spanking
and depressive symptoms, our work suggests that
spanking may be a damaging disciplinary tactic.
Therefore, given the regularity with which American
parents employ spanking (Straus, 2001), it is incum-
bent upon practitioners to educate parents on how
to use other strategies, such as withdrawal of privi-
leges and time outs, which are effective in reducing
behavior problems among children, without the risk
of harming the child’s mental health (Gershoff,
2002; Larzelere & Kuhn, 2005).
What are the implications of our study for the
potential impact of spanking on depressive symp-
toms over time? Because we found no relationship
over time, one could argue that spanking and its
relationship to depressive symptoms is subsumed by
other factors in the long term. For example, other
research has found that spanking and other forms of
physical discipline are related to a complex web of
problematic outcomes including, but not limited to,
increased levels of delinquency, deficits in cognitive
347
skills, and lower self-esteem (Smith & Brooks-
Gunn, 1997; Straus, 1991). Therefore, even without
direct associations in our change model, it appears
that the impact of spanking on youth development
has the potential to be considerable, even in the
absence of a direct association, and should be
avoided (Mulvaney & Mebert, 2007; Straus, 2001).
A second implication of our work is that practi-
tioners should not only pay careful attention to the
disciplinary strategies chosen by parents but also
help parents develop ways to provide emotional sup-
port and warmth for their children. Whether such
support is in terms of positive reinforcement
directed at the children or the adoption of parenting
approaches (e.g., authoritative) that necessitate posi-
tive communication styles, our work shows that the
development of emotional support for the child may
have short- and long-term benefits in reducing
depressive symptoms. Finally, an additional implica-
tion emerges from this work related to the treatment
of depressive symptoms in children and their
mothers. Recall that maternal depressive symptoms
were a salient factor for both African American and
European American youth. Clinical and program-
matic efforts would do well to note the importance
of maternal functioning in the treatment and reduc-
tion of depressive symptoms for youth. Indeed,
other recent research has indicated that a vigorous
approach to the treatment of maternal depression is
warranted and that ‘‘. . . [the] failure to treat
depressed mothers may increase the burden of illness
in their children’’ (Weissman et al., 2006, p. 1397).
For example, many of the treatments (e.g., pharma-
cotherapies) that have proven efficacy in adults
remain quite controversial for children and adoles-
cents. These treatments have been challenged as
ineffective and even dangerous (Weller, Tucker, &
Weller, 2005). Therefore, given how important
maternal depressive symptoms were in the models
developed here, the implementation of maternal
treatment, in addition to direct services to the child,
may be the one of the most effective frontline efforts
to combat the growing prevalence of depression in
youth.
In terms of future research, this study indicates
the need for a greater understanding of which factors
are relevant to child and adolescent depressive symp-
toms by race. The findings indicate that a culturally
competent approach to research must consider pro-
cesses and outcomes by race. For instance, scholar-
ship on race differences in adult depressive
348 Family Relations
Volume 57, Number 3
July 2008
symptoms has pinpointed the disproportionate
number of racial minorities living in poverty, with
low education and few structural opportunities as
the cause of the disparate rates of psychological dis-
tress between Whites and Blacks (Christie-Mizell
et al., 2003; Cutrona, Russell, Hessling, Brown, &
Murry, 2000; Williams, 1990). Noteworthy in the
current study is that socioeconomic differentials do
not appear to account for direct mental health differ-
ences between African Americans and European
Americans. In our final cross-sectional and change
models, not a single socioeconomic indicator was
significantly related to child and adolescent depres-
sion. Nevertheless, some of these factors had bivari-
ate relationships (Table 1) with all our variables of
interest; so future research should investigate the
extent to which the associations of these variables
may be mediated or moderated in our full analysis.
Limitations
Despite the overall strengths of the findings of this
study, there are a few noteworthy limitations. First,
we are restricted to information for mothers and
their children, leaving unanswered the question of
whether spanking administered by fathers is simi-
larly (or differently) related to child depressive
symptoms. Because father-child relationships have
been shown in other research to influence child and
adolescent outcomes (see, e.g., Furstenberg, 1998),
inclusion of data on fathers might further elucidate
the relationship between spanking and child psycho-
logical well-being. A second important limitation
was the restricted age range (i.e., 29 – 37 years) of
the mothers in our sample. Parental age is associated
not only with maturation and different types of
knowledge about parenting but also with the resour-
ces (e.g., financial capital) and types of social control
that parents can exert over their children that help
produce positive outcomes (Parcel & Menaghan,
1994). Therefore, our findings may not generalize to
children born to older or younger mothers. A final
limitation is that our data do not include measures
of role identity or the extent to which the mothers
in our sample feel that they are adequately fulfilling
the maternal role. Such information may have fur-
ther clarified the potential consequences of spanking
and the emotional support on depressive symptoms.
Other research has clearly linked role identity and
role satisfaction to effects on mental health and indi-
vidual motivations (Simon, 1995, 1997).
In conclusion, our findings confirm that spank-
ing and emotional support are important in under-
standing child and adolescent mental health and
that the patterns of association differ by race. Specif-
ically, spanking is positively associated with depres-
sive symptoms for both African Americans and
European Americans but not over time. Emotional
support is associated negatively with depressive
symptoms for both groups but only with the depres-
sive symptoms of European American youth in the
change models. These findings also contribute to the
larger body of work that seeks to understand how
the nexus between parent and child functioning con-
tributes to outcomes for children. Finally, although
we purposely restricted our study to two race/ethnic
groups (i.e., African Americans and European Amer-
icans), future research should incorporate other
groups (e.g., Latinos and Asians) and continue to
explore under what circumstances spanking and
emotional support matter by race and ethnicity.
These future studies should include both intra- and
intergroup comparisons, which may further clarify
the mechanisms that link disciplinary strategies and
children’s mental health outcomes.
References
Achenbach, T. M., & Edelbrock, C. S. (1981). Behavioral problems and
competencies reported by parents of normal and disturbed children aged
four through sixteen. Chicago: University of Chicago Press.
Achenbach, T. M., & Edelbrock, C. S. (1983). Manual for the child behav-
ior checklist and revised child behavior profile. Burlington: University of
Vermont, Department of Psychiatry.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable
distinction in social psychological research: Conceptual, strategic, and
statistical considerations. Journal of Personality & Social Psychology, 51,
1173–1182.
Baumrind, D. (1996). The discipline controversy revisited. Family Rela-
tions: Journal of Applied Family and Child Studies, 45, 405–414.
Bradley, R. H., & Caldwell, B. M. (1980). The relation of home environ-
ment, cognitive competence, and IQ among males and females. Child
Development, 51, 1140–1148.
Bradley, R. H., Caldwell, B. M., & Elardo, R. (1977). Home environment,
social status, and mental test performance. Journal of Educational
Psychology, 69, 697–701.
Caldwell, B., & Bradley, R. (1984). Home observation for measurement of
the environment. Little Rock: University of Arkansas at Little Rock.
Center for Human Resource Research. (2002). NLSY79 Child and Young
adult data user’s guide. Columbus, OH: Ohio State University.
Christie-Mizell, C. A. (2003). Bullying: The consequences of interparental
discord and child’s self-concept. Family Process, 42, 237–252.
Christie-Mizell, C. A. (2004). The immediate and long-term effects of family
income on child and adolescent bullying. Sociological Focus, 37(1), 25–
41.
Christie-Mizell, C. A. (2006). The effects of traditional family and gender
ideology on earnings: Race and gender differences. Journal of Family
and Economic Issues, 27, 48–71.
 Child Depressive Symptoms and Spanking
Christie-Mizell et al.
Christie-Mizell, C. A., Steelman, L. C., & Stewart, J. (2003). Seeing their
surroundings: The effects of neighborhood setting on race and maternal
distress. Social Science Research, 32, 402–428.
Cutrona, C. E., Russell, D. W., Hessling, R. M., Brown, P. A., & Murry,
V. (2000). Direct and moderating effects of community context on the
psychological well-being of African American women. Journal of Person-
ality & Social Psychology, 79, 1088–1101.
Deater-Deckard, K., & Dodge, K. A. (1997). Externalizing behavior prob-
lems and discipline revisited: Nonlinear effects and variation by culture,
context, and gender. Psychological Inquiry, 8, 161–175.
Deater-Deckard, K., Dodge, K. A., Bates, J. E., & Pettit, G. S. (1996).
Physical discipline among African American and European American
mothers: Links to children’s externalizing behaviors. Developmental
Psychology, 32, 1065–1072.
DeVet, K. A. (1997). Parent-adolescent relationships, physical disciplinary
history, and adjustment in adolescents. Family Process, 36, 311–322.
Flynn, C. P. (1994). Differences in attitudes toward corporal punish. Jour-
nal of Marriage and the Family, 56, 314–324.
Flynn, C. P. (1996). Regional differences in spanking experiences and atti-
tudes: A comparison of northeastern and southern college students.
Journal of Family Violence, 11, 59–80.
Furstenberg, F. F. (1998). Social capital and the role of fathers in the fam-
ily. In A. Booth & A. C. Crouter (Eds.), Men in families: When do they
get involved? What difference does it make? (pp. 295–301). Mahwah, NJ:
Lawrence Erlbaum.
Gershoff, E. T. (2002). Corporal punishment by parents and associated
child behaviors and experiences: A meta-analytic and theoretical review.
Psychological Bulletin, 128, 539–579.
Giles-Sims, J., Straus, M. A., & Sugarman, D. B. (1995). Child, maternal,
and family characteristics associated with spanking. Family Relations,
44, 170–176.
Graham, P. J., & Rutter, M. (1968). The reliability and validity of the psy-
chiatric assessment of the child, II: Interview with the parent. British
Journal of Psychiatry, 114, 581–592.
Grogan-Kaylor, A. (2005). Corporal punishment and the growth trajectory
of children’s antisocial behavior. Child Maltreatment, 10, 283–292.
Guo, G., & VanWey, L. K. (1999). Sibship size and intellectual develop-
ment: Is the relationship causal? American Sociological Review, 64,
169–187.
Hoyert, D. L., Kochanek, K. D., & Murphy, S. L. (1999). Deaths: Final
data for 1997. National Vital Statistics Report, 47(19), 1–114.
Kellam, S. G., Branch, J. D., Agrawal, K. C., & Ensminger, M. E. (1975).
Mental health and going to school: The Woodlawn program of assessment,
early intervention, and evaluation. Chicago: University of Chicago.
Kelley, M. L., Power, T. G., & Wimbush, D. D. (1992). Determinants of
disciplinary practices in low-income black mothers. Child Development,
63, 573–582.
Kessler, R. C., McGonagle, K. A., Zhao, S., Nelson, C. B., Hughes, M.,
Eshleman, S., Wittchen, H., & Kendler, K. S., et al. (1994). Lifetime
and 12-months prevalence of DSM-III-R psychiatric disorders in the
United States: Results from the National Comorbidity Survey. Archives
of General Psychiatry, 51, 8–19.
King, G., & Williams, D. R. (1995). Race and health: A multidimensional
approach to African-American health. New York: Oxford University
Press.
Klein, D. N., Lewinsohn, P. M., & Seeley, J. R. (1997). Psychosocial char-
acteristics of adolescents with a past history of dysthymic disorder:
Comparison with adolescents with past histories of major depressive
and non-affective disorders, and never mentally ill controls. Journal of
Affective Disorders, 42, 127–135.
Larzelere, R. E. (2000). Child outcomes of nonabusive and customary phys-
ical punishment by parents: An updated literature review. Clinical Child
and Family Psychology Review, 3, 199–221.
Larzelere, R. E., & Kuhn, B. R. (2005). Comparing child outcomes
of physical punishment and alternative disciplinary tactics: A meta-
analysis. Clinical Child and Family Psychology Review, 8, 1–37.
McCord, J. (1997). On discipline. Psychological Inquiry, 8, 215–217.
McLoyd, V. C. (1990). The impact of economic hardship on black families
and children: Psychological distress, parenting, and socioemotional
development. Child Development, 61, 311–346.
McLoyd, V. C., & Smith, J. (2002). Physical discipline and behavior prob-
lems in African American, European American, and Hispanic children:
349
Emotional support as a moderator. Journal of Marriage and Family, 64,
40–53.
Mizell, C. A., & Steelman, L. C. (2000). All my children: The consequen-
ces of sibling group characteristics on the marital happiness of young
mothers. Journal of Family Issues, 21, 858–867.
Moldenhauer, Z. (2006). Mood disorders. In B. Melnyk & Z. Moldenhauer
(Eds.), The KySS guide to child and adolescent mental health screening:
Early intervention and health promotion. Cherry Hill, NJ: National
Association of Pediatric Nurse Practitioners.
Molnar, B. E., Buka, S. L., Brennan, R. T., Holton, J. K., & Earls, F.
(2003). A multilevel study of neighborhoods and parent-to-child physi-
cal aggression: Results from the project on human development in
Chicago neighborhoods. Child Maltreatment, 8, 84–97.
Mott, F. L. (1998). Patterning of child assessment completion rates in the
NLSY: 1986-1996. Columbus: The Ohio State University, Center for
Human Resource Research.
Mott, F. L. (2004). The utility of the HOME Scale for child development
research in large national longitudinal survey: The National Longitudi-
nal Survey of Youth, 1979 cohort. Parenting: Science and Practice, 4,
261–273.
Mulvaney, M. K., & Mebert, C. J. (2007). Parental corporal punishment
predicts behavior problems in early childhood. Journal of Family Psy-
chology, 21, 389–397.
Nolen-Hoeksema, S., Morrow, J., & Fredrickson, B. (1993). Response
styles and the duration of episodes of depressed mood. Journal of Abnor-
mal Psychology, 102, 20–28.
Parcel, T. L., & Menaghan, E. G. (1993). Family social capital and child-
ren’s behavior problems. Social Psychology Quarterly, 56, 120–135.
Parcel, T. L., & Menaghan, E. G. (1994). Parents’ jobs and children’s lives.
New York: Aldine de Gruyter.
Pearlin, L. I. (1999). The stress process revisited: Reflections on concepts
and their interrelationships. In C. S. Aneshensel & J. C. Phelan (Eds.),
Handbook of the sociology of mental health (pp. 395–415). New York:
Kluwer Academic/Plenum.
Pearlin, L. I., Lieberman, M., Menaghan, E., & Mullan, J. (1981). The
stress process. Journal of Health and Social Behavior, 22, 337–356.
Pearlin, L. I., & McCall, M. E. (1990). Occupational stress and marital
support: A description of microprocesses. In J. Eckenrode & S. Gore
(Eds.), Stress between work and family (pp. 39–60). New York: Plenum.
Peterson, J. L., & Zill, N. (1986). Marital disruption, parent-child relation-
ships, and behavior problems in children. Journal of Marriage and the
Family, 48, 295–307.
Polaha, J., Lazelere, R. E., Shapiro, S. K., & Pettit, G. S. (2004). Physical
discipline and child behavior problems: A study of ethnic group differ-
ences. Parenting: Science and Practice, 4, 339–360.
Radloff, L. (1977). The CES-D scale: A self-report depression scale for
research in the general population. Journal of Applied Psychological
Measurement, 1, 385–401.
Rohner, R. P., Bourque, S. L., & Elordi, C. A. (1996). Children’s percep-
tions of corporal punishment, caretaker acceptance, and psychological
adjustment in a poor, biracial southern community. Journal of Marriage
and the Family, 58, 842–852.
Rutter, M. (Ed.). (1970). Sex differences in children’s responses to family stress.
New York: Wiley.
Sears, R. R., Maccoby, E. E., & Levin, H. (1957). Patterns of child rearing.
New York: Harper & Row.
Shaffer, D., Gould, M. S., Fisher, P., Trautment, P., Moreau, D.,
Kleinman, M., & Flory, M., et al. (1996). Psychiatric diagnosis in child
and adolescent suicide. Archives of General Psychiatry, 53, 339–348.
Simon, R. W. (1995). Gender, multiple roles, role meaning and mental
health. Journal of Health and Social Behavior, 36, 182–194.
Simon, R. (1997). The meaning individuals attach to role identities and
their implications for mental health. Journal of Health and Social Behav-
ior, 38, 256–274.
Smith, J. R., & Brooks-Gunn, J. (1997). Correlates and consequences of
harsh discipline for young children. Archives of Pediatric and Adolescent
Medicine, 151, 777–786.
Stolzenberg, R. M. (1980). The measurement and decomposition of causal
effects in nonlinear and nonadditive models. In K. F. Schuessler (Ed.),
Sociological methodology (pp. 459–488). San Francisco: Jossey-Bass.
Straus, M. A. (1991). Discipline and deviance: Physical punishment of children
and violence and other crime in adulthood. Social Problems, 38, 133–154.
350 Family Relations
Volume 57, Number 3
July 2008
Straus, M. A. (1994). Should the use of corporal punishment by parents be
considered child abuse? In M. A. Mason & E. Gambrill (Eds.), Debat-
ing children’s lives: Current controversies on children and adolescents
(pp. 196–222). Newbury Park, CA: Sage.
Straus, M. A. (2000). Corporal punishment and primary prevention of
physical abuse. Child Abuse and Neglect, 24, 1109–1114.
Straus, M. A. (2001). New evidence for the benefits of never spanking.
Society, 38(6), 52–60.
Straus, M. A., & Stewart, J. H. (1999). Corporal punishment by American
parents: National data on prevalence, chronicity, severity, and duration,
in relation to child and family characteristics. Clinical Child and Family
Psychology Review, 2, 55–70.
Substance Abuse and Mental Health Services Administration. (2007). The
national survey on drug use and health report: State estimates of depression,
2004 and 2005. Rockville, MD: Author.
Turner, H. A., & Muller, P. A. (2004). Long-term effects of child corporal
punishment on depressive symptoms in young adults. Journal of Family
Issues, 25, 761–782.
U.S. Department of Health and Human Services. (1999).
Mental health. A report of the surgeon general. Rockville, MD: U.S.
Department of Health and Human Services, Substance Abuse and
Mental Health Services Administration, Center for Mental Health
Services, National Institutes of Health, National Institute of Mental
Health.
Weissman, M. M., Pilowsky, D. J., Wickramaratne, P. J., Talati, A.,
Wisniewski, S. R., Fava, M., Hughes, C. W., Garber, J., Malloy, E.,
King, C. A., Cerda, G., Sood, A. B., Alpert, J. E., Trivedi, M. H., &
Rush, A. J. (2006). Remissions in maternal depression and child psy-
chopathology—A star*d-child report. Journal of the American Medical
Association, 295, 1389–1398.
Weissman, M. M., Scholomskas, D., Pottenger, M., Prusoff, B. A., &
Locke, B. Z. (1977). Assessing the depressive symptoms in five psychi-
atric populations: A validation study. American Journal of Epidemiology,
106, 206–214.
Weissman, M. M., Wolk, S., Wickramaratne, P., Goldstein, R. B.,
Adams, P., Greenwald, S., Ryan, N. D., Dahl, R. E., & Stemberg, D.
(1999). Children with prepubertal-onset major depressive disorder
and anxiety grown up. Archives of General Psychiatry, 56, 794–
801.
Weller, E. B., Tucker, S., & Weller, R. A. (2005). The selective serotonin
reuptake inhibitors controversy in the treatment of depression in chil-
dren. Current Psychiatry Reports, 7, 87–90.
Williams, D. R. (1990). Socioeconomic differentials in health: A review
and redirections. Social Psychology Quarterly, 53, 81–99.
American Journal of Public Health, Vol. 84, Issue 4 612-617, Copyright © 1994 by American Public Health
Association
Factors associated with the use of violence among urban
black adolescents.
R H DuRant, C Cadenhead, R A Pendergrast, G Slavens and C W Linder

Division of Adolescent/Young Adult Medicine, Children's Hospital, Harvard Medical School, Boston, MA 02115.

OBJECTIVES. The purpose of this study was to examine social and psychological factors
associated with the use and nonuse of violence among Black adolescents living in a community
with a high level of violent crime. METHODS. Adolescents (n = 225, 44% male) 11 to 19 years
of age living in or around nine housing projects in an urban area were administered an
anonymous questionnaire. RESULTS. Self-reported use of violence was associated with
exposure to violence and personal victimization, hopelessness, depression, family conflict,
previous corporal punishment, purpose in life, self-assessment of the probability of being alive at
age 25, and age and was higher among males. CONCLUSIONS. These data support the
hypothesis that exposure to violence is associated with adolescents' self-reported use of violence.
However, adolescents with a higher sense of purpose in life and less depression were better able
to withstand the influence of exposure to violence in the home and in the community.
 Child Development, November/December 2001, Volume 72, Number 6, Pages 1868–1886

The Home Environments of Children in the United States

Part II: Relations with Behavioral Development through Age Thirteen
Robert H. Bradley, Robert F. Corwyn, Margaret Burchinal, Harriette Pipes McAdoo,
and Cynthia García Coll

This study examined the frequency with which children were exposed to various parental actions, materials,
events, and conditions as part of their home environments, and how those exposures related to their well-being.
Part 1 focused on variations by age, ethnicity, and poverty status. In Part 2 of the study, relations between major
aspects of the home environment (including maternal responsiveness, learning stimulation, and spanking) and
developmental outcomes for children from birth through age 13 were investigated. The outcomes examined
were early motor and social development, vocabulary development, achievement, and behavior problems.
These relations were examined in both poor and nonpoor European American, African American, and Hispanic
American families using hierarchical linear modeling. The most consistent relations found were those between
learning stimulation and children’s developmental status, with relations for parental responsiveness and spank-
ing varying as a function of outcome, age, ethnicity, and poverty status. The evidence indicated slightly stronger
relations for younger as compared with older children.

INTRODUCTION ple, it is widely acknowledged that different ethnic

For more than half a century educators and develop-
mentalists have examined relations between children’s
home environments and their behavioral develop-
ment. During this period, a large canon of studies has
accumulated and several excellent reviews have been
published (Bornstein, 1995b; Bradley & Tedesco, 1982;
Collins, Harris, & Susman, 1995; Edwards, 1995; Ford
& Lerner, 1992; Holmbeck, Paikoff, & Brooks-Gunn,
1995; Kagan, 1984; Parke, 1978; Wachs, 1992, 2000).
Even so, constructing a comprehensive set of general-
izations about how the home environment relates to
developmental course across childhood is difficult,
due to the diversity of measures used to assess the
same ostensible constructs and the complexity of the re-
lations themselves. Very rarely, for example, have the
same environmental measures and the same develop-
mental measures been used to examine environment/
development relations over more than two develop-
mental periods. Thus, there remain gaps and incon-
sistencies in what is known.
Current developmental systems theories postulate
that children living in different ecological niches tend
to encounter specific actions, objects, events, and con-
ditions in different amounts and patterns (Bronfen-
brenner, 1995; Chao, 1994; García Coll & Magnusson,
1999; Harkness & Super, 1995; McLoyd, 1999; Wachs,
1992). Such were the findings relative to ethnicity and
poverty status in Part 1 (Bradley, Corwyn, McAdoo,
& García Coll, 2001) of this report. The theories also
postulate that children living in different ecologies at-
tach different meanings to these encounters. For exam-
groups have different views of appropriate childrear-
ing and different developmental goals for their chil-
dren (Greenfield, 1995). Deater-Deckard, Dodge, Bates,
and Pettit (1996) suggest that harsh discipline may im-
ply an out-of control, parent-centered household for
European Americans, whereas lack of physical disci-
pline among African Americans may indicate an abdi-
cation of the parenting role. Thus, the impact of specific
features of the environment on children’s competence
and behavior is likely to vary across ecological niches.
The complexity of most ecologies and the multidimen-
sionality of most ecological/developmental models
make it difficult to predict with precision what the rela-
tions are likely to be for most developmental ecologies.
For example, in the United States, ethnicity is often
confounded with socioeconomic status (SES). It is
generally acknowledged that economic hardship has
a negative impact on development (Bolger, Patterson,
Thompson, & Kupersmidt, 1995; Duncan, Brooks-
Gunn, & Klebanov, 1994), but much remains uncertain
about how ethnicity and economic status combine to
affect the pattern of relations between aspects of chil-
dren’s environments and the course of development—
although, several recent studies are helping to fill in
some gaps (Brody & Flor, 1998; Florsheim, Tolan, &
Gorman-Smith, 1998; Pettit, Bates, & Dodge, 1997).
Moreover, ethnicity and SES are confounded with re-
ligion, area of residence, family configuration, and so-
© 2001 by the Society for Research in Child Development, Inc.
All rights reserved. 0009-3920/2001/7206-0016

cial and political history, all of which presumably help
determine what children experience and how those
experiences operate to influence children’s develop-
ment. The influence of family is both circumscribed
and enhanced in terms of these setting conditions and
what they portend in terms of access to opportunities
and resources in nonfamily contexts (Bronfenbrenner,
1995; McLoyd, Cauce, Takeuchi, & Wilson, 2000;
Steinberg, Darling, Fletcher, Brown, & Dornbusch,
1995). Relatedly, as Greenfield (1994) has argued,
“Culture not only is context: it has context as well” (p.
25). The consequences of any experience that children
from a particular ethnic group have in their own
home environments depends on the history of inter-
group relations in that group’s current societal con-
text and the degree of connectedness between family
life and other societal institutions designed to sup-
port child well-being. In effect, the broader society is
less likely to provide the support and more likely to
present interferences with the developmental scripts
established for children of color (García Coll & Mag-
nusson, 1999; Scott-Jones, 1995). Because there is less
consistency across contexts in how aspects of the en-
vironment are enacted and consequated for children
of color, there is less likelihood that aspects of the
home environment will exert a powerful effect on de-
velopmental course (Bloom, 1964).
The issue of environmental influence is further
complicated by changes in how children engage their
environments during different age periods. Interac-
tions with parents become less frequent as children
grow older, whereas interactions with others (most
notably peers) become more frequent (Collins et al.,
1995). Thus, there is reason to suspect that certain as-
pects of the home environment may be less strongly
associated with child behavior and development in
older children. Relatedly, as Scarr and McCartney
(1983) have argued, when children are young, they
are primarily in the role of reactors to and elicitors of
specific environmental inputs. As they age, children
are far more active in seeking out and constructing
environments that suit their needs and proclivities.
Indeed, estimates of the heritability of many charac-
teristics increase with age, suggestive of more active
selection of environments deemed desirable by chil-
dren as they grow older (Scarr, 1992). The timing of this
evolution in style of engagement with the environment
is itself at least somewhat contextually conscripted.
Children from some ecological niches are afforded
greater access to certain out-of-home environments,
and children from some ecological niches have greater
incentive to engage certain out-of-home environments.
Bradley et al. 1869
In either case, the time spent in the home could be less-
ened, as might the influence exerted by the home en-
vironment. Relatedly, Scarr (1992) and others (e.g.,
Rowe, Jacobson, & Van den Oord, 1999) have argued
that environmental influences are greatest in condi-
tions that pose a risk or hazard for development. Re-
sults from relatively simple correlational analyses pro-
vide mixed support for this proposition (e.g., Bradley
et al., 1989). However, Rowe and colleagues (1999) re-
cently found that environmental conditions shared
by members of the same family had greater influence
in families in which mothers had low educational at-
tainment as compared with families in which mothers
had high educational attainment. Data from the Na-
tional Longitudinal Survey of Youth (NLSY) affords
an opportunity to examine these patterns in several
major ethnic groups and for children from different
economic strata using a common set of home environ-
ment indicators and a common set of developmental
measures.
In the present study (Part 2 of this report) we inves-
tigated aspects of the home environment that previ-
ous research has shown to be associated with chil-
dren’s well-being and that developmental theorists
postulate as mechanisms that contribute to individ-
ual differences in behavioral development (Bradley &
Caldwell, 1995): learning stimulation (both objects and
experiences), parental responsiveness, and spanking.
For specific developmental epochs other aspects of the
home environment were examined as well. These asso-
ciations were studied both cross-sectionally and longi-
tudinally for three major ethnic groups: European
Americans, African Americans, and Hispanic Ameri-
cans. The aspects of the home environment were exam-
ined both separately and in specific combinations in an
effort to describe the patterns of association as fully as
possible with the data available. Our primary intent
was to describe the pattern of associations within each
ethnic and economic group to provide for future in-
vestigations a kind of information backdrop of these
relations using more detailed and comprehensive
measures of the home environment and children’s
developmental functioning. Although our intent was
to discuss similarities in patterns of relations across
groups, we did not examine group differences per se.
METHODS
Sample
Data for this study came from five biennial NLSY
child data files from 1986 through 1994 (see Center for
1870 Child Development
Human Resource Research [CHHR], 1995, 1997b). The
sample was the same as that used for Part 1 of the
study (Bradley et al., 2001) with one exception: chil-
dren over the age of 13 were not included, due to lim-
itations in the data available for older children.
Measures
Home Observation for Measurement of the Environment–
Short Form. The home environment was measured
with the Home Observation for Measurement of the
Environment–Short Form (HOME-SF; Baker, Keck,
Mott, & Quinlan, 1993). Extensive information about
the HOME-SF can be found in Part 1 of the study
(Bradley et al., 2001). There were four age-appropriate
versions of the HOME-SF: the Infant–Toddler HOME
(IT-HOME) for children under age 3; the Early Child-
hood HOME (EC-HOME) for children between the
ages of 3 and 5; the Middle Childhood HOME (MC-
HOME) for children ages 6 through 9; and the Early
Adolescent HOME (EA-HOME) for children 10 and
over. The same five home environment factors exam-
ined in Part 1 of the study were examined in Part 2;
specifically, learning stimulation, parental responsive-
ness, spanking, teaching, and physical environment.
The first three were used in longitudinal analyses (hi-
erarchical linear modeling [HLM] described below)
because they were available at all assessment points
from birth through age 13. The remaining two were
used in correlational analyses (also described below).
General development during infancy. For children from
birth through 48 months of age, mothers completed
the NLSY Motor and Social Development assessment
(Baker et al., 1993) developed by the National Center
for Health Statistics. Items were derived from standard
measures of child development (e.g., Bayley, Gesell, &
Denver; Baker et al., 1993). Original technical analyses
of this measure were based on data from, 2714 chil-
dren who participated in the 1981 Child Health Sup-
plement to the National Health Interview Survey. Ex-
tensive analysis of this measure at the CHHR of Ohio
State University indicated a very low noncompletion
rate (about 6%); little evidence of ethnic, gender, or
education bias; and meaningful correlations with
other measures taken during infancy. The informa-
tion used to score the items on this measure, however,
were all obtained via maternal report; thus, the mea-
sure has limitations as an assessment of children’s
early motor and social development.
Behavior problems index. For children aged 4 and
older, mothers completed the 28-item Behavior Prob-
lem Index (BPI; Baker et al., 1993). The scale is de-
signed to measure the frequency, range, and type of
behavior problems. It was developed by Zill and
Peterson of Child Trends, Inc. for use in the National
Health Interview Survey. Items were derived from the
Achenbach Child Behavior Checklist and other com-
monly used child behavior scales. Factor analysis of
the 28 items indicated that the scale assesses six areas:
antisocial, anxious/depressed, headstrong, hyperac-
tive, immature dependency, and peer conflict/social
withdrawal. The  was estimated at .89 for young chil-
dren, and .91 for adolescents. The test–retest correla-
tion, corrected using the Spearman-Brown formula,
was estimated at .92.
Peabody Individual Achievement Test. Children age 5
years and older were administered three subtests
from the Peabody Individual Achievement Test (PIAT;
Dunn & Markwardt, 1970) battery: mathematics, read-
ing comprehension, and reading recognition. Because
of technical problems with the reading comprehension
subtest for younger children and the fact that the two
reading subtests were so highly intercorrelated, how-
ever, only the math and reading recognition subtests
were analyzed for this report. The PIAT was standard-
ized on 2,887 children in kindergarten through grade
12 in the late 1960s. One-month test–retest correlation
was estimated at .74 for the math subtest. Completion
rates for this subtest ranged from about 89% for His-
panic American children to 94% for European Ameri-
can and African American children. The test–retest
correlation for the reading recognition subtest was .89.
Completion rates for this subtest were similar to those
for the math subtest. The lower rate of completion for
Hispanic American children was anticipated, given
that the test was only administered in English.
Peabody Picture Vocabulary Test – Revised. The Pea-
body Picture Vocabulary Test–Revised (PPVT-R; Dunn
& Dunn, 1981) measures receptive vocabulary for stan-
dard American English. The PPVT-R was given to all
children ages 3 and older, and was administered at the
first possible assessment point (ages 3 to 4 for most chil-
dren, but at later ages for children who were older by the
first year that child assessments were done in 1986). A
renewed effort was made to obtain PPVT-R assessments
when children were between 10 and 12 years old. Thus,
only the data from two age points were used (3- to
4-year-olds and 10- to 11-year-olds), except for the HLM
analyses, which included PPVT-R scores from interven-
ing ages as well. The completion rates were 90% for Eu-
ropean American children, 86% for African American
children, and 81% for Hispanic American children.
Poverty status. The determination of family pov-
erty status (i.e., “in poverty” versus “not in poverty”)
was based on CHHR-created estimates in 1986, and
the official poverty income guidelines issued by the
Department of Health and Human Services (DHHS)
for the years 1988 through 1994 (CHHR, 1997a). Com-
 Bradley et al. 1871

parable to the DHHS guidelines, the CHHR estimates
were determined by total family income, controlling
for family size, farm/nonfarm residence, and state of
residence. These poverty levels were compared with
the mother’s reported income, below which the fam-
ily was considered in poverty for that year.
Procedure
The ages at which each measure were given and
the fact that data were collected only every other year
helped to determine the structure of the data analy-
ses. Specifically, (1) the IT-HOME-SF was applicable
to children age birth to 3, the EC-HOME for ages 3 to
5, the MC-HOME for ages 6 to 9, and the EA-HOME
for ages 10 to 14; (2) the Motor and Social Develop-
ment assessment was applicable from birth through
age 3; (3) the PPVT-R was used from age 3 onward; (4)
the PIAT was used from age 5 onward; and (5) the BPI
was available from age 4 onward. As a result, analy-
ses were organized according to the following age
groups, with about half of the sample included in
the 1-year age groups and all of the sample included
in the 2-year age groups: birth to 11 months, 12 to
35 months, 36 to 59 months, 60 to 71 months, 72 to 95
months, 96 to 119 months, 120 to 143 months, and 144
to 167 months. Means and standard deviations for all
child outcome measures in all groups at every age an-
alyzed are presented in Table 1.
RESULTS
Data Analysis Strategy
The primary statistical technique used to examine
relations between children’s home environments and
aspects of their intellectual and social development
was HLM (Bryk & Raudenbush, 1992). HLM analyses
were conducted for five child developmental measures
(early motor and social development based on items
from the Bayley Scales, PPVT-R, BPI, PIAT math, and
PIAT reading recognition). HLM was used to take ad-

Table 1 Means of Child Outcome Variables at Different Ages

European American African American Hispanic American

Variable Age (yr) Poor Nonpoor Poor Nonpoor Poor Nonpoor

Social and motor 0 99.92 (14.08) 101.08 (13.75) 103.96 (13.71) 100.69 (14.81) 98.81 (13.82) 102.72 (11.56)
1 and 2 103.44 (13.60) 102.11 (14.44) 103.89 (14.53) 101.87 (15.29) 98.27 (14.23) 97.28 (14.39)
3 101.49 (12.60) 94.62 (14.56) 96.77 (12.31) 92.87 (13.08) 92.50 (13.26) 91.52 (14.22)
Peabody Picture 3 and 4 98.60 (15.78) 87.70 (17.59) 78.07 (15.95) 71.83 (15.13) 81.19 (17.75) 72.91 (17.46)
Vocabulary–Revised 5 98.30 (14.31) 93.19 (14.32) 84.05 (15.05) 77.57 (14.46) 83.56 (17.24) 75.76 (16.87)
6 and 7 100.24 (15.49) 95.12 (16.76) 85.25 (14.95) 79.19 (14.82) 86.01 (18.01) 76.60 (16.23)
8 and 9 99.42 (14.06) 92.21 (16.25) 84.94 (16.53) 77.82 (14.78) 90.69 (19.88) 75.13 (17.89)
10 and 11 100.01 (14.27) 92.02 (16.00) 86.57 (15.51) 78.90 (16.20) 87.22 (16.98) 79.51 (15.94)
12 and 13 97.08 (14.02) 92.89 (17.42) 86.43 (12.92) 78.11 (16.64) 90.79 (16.59) 83.00 (22.24)
Behavior Problems Index 4 104.46 (14.31) 109.83 (15.04) 104.84 (13.59) 108.68 (16.42) 102.86 (13.52) 111.20 (15.18)
5 104.36 (14.04) 109.53 (14.99) 104.89 (13.58) 109.35 (15.93) 106.46 (13.89) 109.33 (14.52)
6 and 7 105.71 (14.04) 110.01 (14.72) 106.49 (12.31) 109.16 (14.46) 105.55 (14.56) 108.64 (15.40)
8 and 9 107.77 (13.93) 112.37 (14.98) 106.47 (13.73) 110.11 (14.51) 107.11 (13.86) 110.75 (14.93)
10 and 11 108.41 (14.05) 112.23 (15.14) 106.13 (13.27) 111.12 106.83 (13.49) 110.82 (14.64)
12 and 13 110.14 (13.78 107.12 (14.69) 108.19 (14.12) 111.77 (15.96) 106.97 (13.51) 113.33 (14.93)
Peabody Individual 5 102.64 (13.98) 96.22 (14.55) 96.47 (14.35) 90.74 (13.69) 96.06 )12.63) 89.68 (15.71)
Achievement Test—math 6 and 7 103.13 (10.87) 99.24 (10.87) 98.64 (11.05) 95.02 (11.53) 97.81 (11.44) 93.71 (11.34)
8 and 9 103.08 (12.00) 98.49 (13.63) 95.77 (12.38) 92.18 (11.84) 98.35 (12.23) 92.91 (11.52)
10 and 11 102.55 (12.54) 96.90 (12.91) 95.70 (12.96) 91.87 (13.07) 96.40 (13.50) 91.52 (12.53)
12 and 13 101.68 (11.69) 96.51 (12.04) 95.51 (11.05) 91.26 (11.51) 96.76 (12.63) 91.20 (11.57)
Peabody Individual 5 109.21 (13.94) 102.60 (13.77) 109.33 (14.25) 101.69 (14.29) 102.75 (14.56) 94.04 (12.78)
Achievement Test— 6 and 7 104.76 (11.24) 100.72 (11.66) 103.98 (10.54) 100.21 (10.73) 101.99 (11.48) 96.69 (10.68)
reading recognition 8 and 9 107.27 (13.20) 101.41 (14.81) 102.09 (14.29) 97.54 (13.52) 103.34 (13.85) 95.74 (15.69)
10 and 11 106.01 (13.81) 100.64 (13.81) 99.72 (13.86) 95.01 (13.54) 101.76 (14.28) 96.40 (14.45)
12 and 13 103.48 (14.18) 98.82 (14.14) 98.45 (13.38) 93.01 (14.40) 102.59 (15.27) 95.46 (16.10)

Note: Values in parentheses represent standard deviations. In order to save space, Table 1 does not contain the full output for hierarchical
linear modeling (HLM) analyses. The table only contains results for the Home Observation for Measurement of the Environment–Short
Form (HOME-SF) and Age  HOME-SF effects for the six major Ethnicity  Poverty Status subgroups. Other significant effects from the
HLM analyses are reported in the text.
1872 Child Development
vantage of the longitudinal structure of the data. Spe-
cifically, the HLM analyses were conducted on those
participants for whom the same outcome measures
were available across time. For these analyses individ-
ual growth curves for a given dependent variable (e.g.,
were first estimated PPVT-R scores). Unfortunately, be-
cause so few children had more than two repeated as-
sessments for most developmental outcomes, it was
not feasible to estimate individual rates of change.
The second stage of analysis involved using indi-
vidual trajectories, or growth curves, as outcome vari-
ables in a between-persons analysis. For each analysis,
child age, HOME-SF scores, ethnicity, poverty status,
number of household residents, maternal education,
and father presence were treated as covariates in the
model. Although the environments that parents pro-
vide for their children reflect aspects of their personality
(e.g., aptitude, intelligence, mental health, and self-
esteem) as well as the content of their parenting, we
decided not to use personality measures as covariates
for several reasons: (1) the NLSY dataset included only
limited measures of mothers’ personality and no data
on fathers’ personality; (2) the data that was available,
such as the Armed Forces Qualifications Test for
mothers, was highly correlated with other covariates
in the model (e.g., maternal education); and (3) there
is debate on whether some of the personality mea-
sures used are equally valid in all cultural and socio-
economic groups.
Three HOME-SF scores were available on children
at all age points: learning stimulation, parental respon-
siveness, and spanking. Thus, these three scores were
used in the HLM analyses. We assumed that scores on
these three factors represented the same constructs
across assessment points, despite some variation in
the item content for learning stimulation and parental
responsiveness. We made this assumption because
those two constructs correspond to the organization
of the original HOME scale (Caldwell & Bradley, 1984)
and because those two item groupings consistently
emerge in factor analyses performed on the HOME
scale (Bradley, 1994). Population curves were estimated
for the six Poverty Status  Ethnic Group combinations
(poor and nonpoor crossed with African American,
Hispanic American, and European American). These
six population curves described the average pattern of
change over time associated with a particular HOME-
SF factor scale for each Poverty Status  Ethnic Group
combination, controlling for the other two HOME-SF
factors and demographics. For example, separate pop-
ulation curves were estimated that described the asso-
ciation between learning stimulation and PIAT math
for poor African Americans, nonpoor African Ameri-
cans, poor European Americans, nonpoor European
Americans, poor Hispanic Americans, and nonpoor
Hispanic Americans. All covariates, including child
age, were centered so that the intercept represented
predicted values at the mean child age, mother’s edu-
cation, and household size.
The between-subjects component of the HLM anal-
yses included main effects plus interactions. Our ap-
proach to analysis involved testing the main effects
and interaction effects as a block first, followed by an
examination of each effect separately. The main ef-
fects included child age, ethnicity (African American,
Hispanic American, European American) and poverty
status (poor, nonpoor), and the covariates maternal
education, presence of a father, and household size.
Interactions tested included (1) all possible interac-
tions between ethnic group membership and poverty
group membership; (2) the interaction between the
three HOME-SF scales and ethnic group membership;
(3) the interaction between the three HOME-SF scales
and poverty group membership; (4) the interaction be-
tween age and ethnic group membership; (5) the inter-
action between age and poverty group membership;
(6) the interaction between age and the three HOME-
SF scales; (7) the interaction among age, poverty group
membership, and the three HOME-SF scales; (8) the in-
teraction among age, ethnic group membership, and
the three HOME-SF scales; and (9) the interaction
among age, ethnic group membership, poverty group
membership, and the three HOME-SF scales. The block
tests displayed in Tables 2 through 6 were conducted
as follows: (1) a block test for the main effect of the
three HOME-SF scales (the F value for this effect can
be found in the first row, labeled “HOME-SF scales,”
under the column heading “HOME” in Tables 2–6);
(2) the block test of the interaction between HOME-SF
and ethnic group membership (the F value for this effect
can be found in the first row, labeled “HOME-SF
Scales,” under the column heading “HOME  Eth-
nicity”); (3) the block test of the interaction between
HOME-SF and poverty group membership (the F
value for this effect can be found in the first row, la-
beled “HOME-SF scales,” under the column heading
“HOME  Poverty”); (4) the block test of the three-
way interaction among HOME-SF, poverty group
membership, and ethnic group membership (the F
value for this effect can be found in the first row, la-
beled “HOME-SF Scales,” under the column heading
“HOME  Poverty  Ethnicity”); (5) the block test
of the interaction between child age and the three
HOME-SF scales (the F value for this effect can be
found in the row labeled “Age  HOME-SF” under
the column heading “HOME”); (6) the block test of the
three-way interaction among child age, HOME-SF,
and poverty group membership (the F value for this

effect can be found in the row labeled “Age  HOME-
SF” under the column heading “HOME  Poverty”);
(7) the block test of the three-way interaction among
child age, HOME-SF, and ethnic group membership
(the F value for this effect can be found in the row
labeled “Age  HOME-SF” under the column heading
“HOME  Ethnicity”); and (8) the four-way interac-
tion among child age, HOME-SF, poverty group mem-
bership, and ethnic group membership (the F value for
this effect can be found in the row labeled “Age 
HOME-SF” under the column heading “HOME  Eth-
nicity  Poverty”). For each block test that was signif-
icant, p  .05, the test statistic for each of the three
HOME-SF factors is listed separately in the rows corre-
sponding to their labels.
We did not perform HLM analyses for the other two
HOME-SF factors examined in Part 1, teaching and the
physical environment, because teaching scores were
only available for children ages 3 to 5, and physical en-
vironment scores were only available for children ages
3 to 9. Partial correlations were computed between
these two factors and children’s performance on the
BPI, PPVT-R, PIAT reading, and PIAT math at every
age point when the factor scores were available. In
each analysis we controlled for father presence, num-
ber of household residents, and maternal education.
Six sets of partial correlations were performed for
each HOME-SF factor (teaching, physical environ-
ment) and each developmental outcome (BPI, PPVT-
R, PIAT reading, PIAT math) combination, one for
each Poverty  Ethnicity subgroup. Results of these
analyses are included in Table 7.
To reiterate, we decided not to analyze data on
14-year-olds because the data from the NLSY are less
representative at that age—data on children age 10
and older are not as representative as data on younger
children, but we decided to use them to take as much
advantage as possible of the NLSY sample. Because of
the very large number of analyses completed for this
study, results are described succinctly. Tables display-
ing the full set of results, including simple bivariate
analyses, can be obtained from the first author. (The
full set of tables can be obtained by accessing our Web
site: http://ualr.edu/home.)
Prior to reporting on the findings, it is probably
worth describing the stability of key HOME-SF and
child developmental measures. After the first year of
life, the amount of learning stimulation available to
children remains moderately stable through age 13;
the stability coefficients were typically in the .5 to .6
range for data gathered 2 years apart. By comparison,
scores for parental responsiveness and spanking were
not nearly as stable. The stability coefficients for pa-
rental responsiveness were about .2 through age 9, in-
Bradley et al. 1873
creasing to about .4 in adolescence. For spanking,
scores were quite unstable through age 3; there was
modest stability from age 3 to 5 (about .3), followed by
instability to age 9, and modest stability again during
the transition to adolescence (about .3). The apparent
lack of stability for both parental responsiveness and
spanking may partially reflect the small number of in-
dicators of each construct available in the NLSY
dataset. Research on parental responsiveness often in-
volves the use of more extensive and labor-intensive
measurement procedures (e.g., for research on respon-
siveness, see Bornstein, 1989). Previous studies on
spanking and other forms of harsh punishment, how-
ever, indicate only low to moderate stability (Strauss,
1991). BPI scores showed substantial stability from age
4 through age 13 (about .6). PIAT reading recognition
scores showed substantial stability after age 8 (about .6
to .7), whereas PIAT math scores were stable from age
5 onward (.5 to .7). Likewise, PPVT-R scores were sta-
ble from age 3 onward (.5 to .7). For both the PIAT and
the PPVT-R, the scores were slightly more stable after
school entry.
Early Motor and Social Development
Table 2 displays the HLM results for early motor
and social development. The block test for the three
HOME-SF factors was significant, F(3, 3865)  56.2,
p  .001. Both learning materials, overall   11.61,
F(1, 3878)  147.9, p  .001, and parental responsive-
ness, overall   3.26, F(1, 3842)  8.79, p  .01,
showed low to moderate relations with early motor
and social development. The overall tests for interac-
tions among the HOME-SF factors, ethnic group
membership, poverty group membership were not
significant. Spanking showed little relation to early
motor and social development.
PPVT-R
Results from the HLM analysis of PPVT-R scores are
shown in Table 3. Results displayed in the top half of
the table pertain to the effects of the HOME-SF, poverty
group membership, and ethnic group membership,
across age. Results in the bottom half of the table per-
tain to interactions between age and the HOME-SF.
The main effect for each HOME-SF factor is estimated
as the mean of the coefficients for the six Ethnicity 
Poverty Status subgroups for that factor (e.g., the six
 coefficients in the row associated with the HOME-
SF factor, learning stimulation). Likewise, Age 
HOME-SF interaction effects for a particular HOME-
SF factor (e.g., Parental Responsiveness) are esti-
mated as the mean of the coefficients associated with
1874 Child Development

Table 2 HLM Results: Early Motor and Social Development

Regression Coefficients for Each Ethnicity  Poverty Status Group

F Value for Total Sample
European African Hispanic
American American American HOME 
HOME  HOME  Ethnicity 
Poor Nonpoor Poor Nonpoor Poor Nonpoor Overall  HOME Ethnicity Poverty Poverty

HOME-SF scales 56.2*** 2.01 1.58 1.22

Learning 18.38*** 11.23*** 12.83*** 11.02*** 7.83* 8.35*** 11.64*** 147.9***
stimulation (2.58) (1.46) (1.89) (1.95) (3.18) (2.36) (.95)
Parental 2.95 3.41* 6.23** 5.15* 2.07 3.91 3.26** 8.79**
responsiveness (2.79) (1.40) (2.11) (2.31) (3.94) (2.86) (1.10)
Spanking .10 .06 .01 .04 .58 .36 .05 .26
(.20) (.10) (.10) (.12) (.49) (.30) (.11)
Age  HOME-SF scales 1.53 1.18 .22 .62
Learning 2.43 1.57 1.68 1.23 3.65 .86 1.35
stimulation (2.36) (1.07) (1.81) (1.51) (2.81) (1.88) (.81)
Parental 1.63 .09 1.55 1.30 4.34 3.71
responsiveness (2.57) (1.23) (1.83) (1.96) (3.29) (2.28)
Spanking .14 .09 .20 .00 .07 .19 .05
(.24) (.10) (.15) (.10) (.51) (.28) (.11)

Note: Values in parentheses represent standard errors. HLM  hierarchical linear modeling; HOME-SF  Home Observation for Measure-
ment of the Environment–Short Form.
* p  .05; ** p  .01; *** p  .001

Table 3 HLM Results: Peabody Picture Vocabulary Test–Revised

Regression Coefficients for each Ethnicity  Poverty Status Group

F Value for Total Sample
European African Hispanic
American American American HOME 
HOME  HOME  Ethnicity 
Poor Nonpoor Poor Nonpoor Poor Nonpoor Overall  HOME Ethnicity Poverty Poverty

HOME-SF scales 43.9*** .90 .55 2.15*

Learning 9.83*** 9.60*** 7.13*** 6.39*** 4.86 11.72*** 8.26*** 81.4*** 1.63
stimulation (2.58) (1.37) (1.70) (1.83) (2.97) (2.19) (.91)
Parental 2.15 4.22*** 2.47 3.38* 7.24** 3.75 3.86*** 29.8*** 1.19
responsiveness (1.87) (1.14) (1.35) (1.41) (2.28) (1.93) (.71)
Spanking .38 .31 .15 .73*** .65 .35 .31 6.42* 3.78*
(.25) (.26) (.10) (.21) (.48) (.33) (.12) a*
Age  HOME-SF scales 2.31 .75 1.83 1.05
Learning 1.56 .41 .03 .17 2.05* .59 .61* 4.11*
stimulation (.91) (.45) (.55) (.58) (.99) (.68) (.30)
Parental .36 .94** .40 .25 .48 .27
responsiveness (.60) (.32) (.40) (.40) (.72) (.52)
Spanking .02 .02 .02 .07 .01 .00 .01
(.08) (.08) (.04) (.05) (.16) (.12) (.04)

Note: Values in parentheses represent standard errors. HLM  hierarchical linear modeling; HOME-SF  Home Observation for Measure-
ment of the Environment–Short Form.
a  significant third-order interaction involving African Americans versus Hispanic Americans versus poverty versus nonpoverty.
* p  .05; ** p  .01; *** p  .001.

the six Ethnicity  Poverty Status subgroups. The co-
efficients for each HOME-SF factor score represent
the estimated increment in PPVT-R associated with a
1-point increase in that HOME-SF score. Given that
the first two factor scores (learning stimulation and pa-
rental responsiveness) reflect the proportion of items
passed, the coefficients displayed in Table 3 for a group
indicate the expected difference in PPVT-R performance
for a child living in a home in which none of the items
were passed and a child living in a home in which all
the items were passed. By contrast, because the spank-
ing score is the number of times a child was spanked in
the past week, the coefficient associated with spanking
represents the estimated change in a child’s PPVT-R
score associated with each additional spanking.
As Table 3 shows, the block test for HOME-SF
scores was significant, F(3, 5227)  43.9, p  .001. The
tests for each of the three HOME-SF factors were also
significant. Additionally, there was a significant three-
way interaction among HOME-SF, poverty group mem-
bership, and ethnic group membership, F(6, 5037)  2.15,
p  .05. Post hoc analyses showed that the interaction
pertained to differences in the effects of spanking for
poor versus nonpoor Hispanic Americans versus Af-
rican Americans. There was no significant interaction
between HOME-SF and age, even though there was
some evidence of a stronger association at younger
ages among impoverished Hispanic American chil-
dren,   2.05.
Learning stimulation was related to PPVT-R scores,
even with maternal education, family poverty status,
father presence, number of household residents, and
ethnicity controlled, F(1, 5427)  81.4, p  .001. Chil-
dren tended to have higher PPVT scores if their fami-
lies scored higher on the learning stimulation HOME-
SF scale (e.g.,   7.13 for African American poor
children indicated that an increase of 1 point on the
HOME-SF was associated, on average, with an in-
crease of 7.13 points on the child’s PPVT score). Al-
though the main effect HOME coefficients were not
always statistically significant—  4.86, p  .05 for
Hispanic American poor children, but   11.72, p 
.001 for Hispanic American nonpoor children—the
differences among the Ethnicity  Poverty Group
coefficients was not reliably different. The results,
therefore, indicated that all children tended to have
higher language scores when the family environment
provided more learning stimulation, overall   8.26,
p  .001 (i.e., an expected difference of 8.26 points be-
tween children from homes in which none and all of the
learning stimulation items were passed). The overall
block test between HOME-SF and age was not signifi-
cant. However, the Learning Stimulation  Age inter-
action was significant, F(1, 3477)  4.11, p  .05,  
Bradley et al. 1875
.61, suggesting a slightly stronger effect for younger
children.
Results from the HLM analysis showed that paren-
tal responsiveness was also associated with PPVT-R
scores, F(1, 4978)  29.8, p  .001. As with learning
stimulation, there were no consistent age differences
in the strength of the association and no significant
interactions with either poverty status or ethnicity.
Again, this analysis suggests that all children tended
to have better language scores if their mothers or pri-
mary caregivers were viewed as more responsive,
overall   3.86, p  .001.
In the HLM analysis, spanking also showed signif-
icant associations with PPVT-R scores, F(1, 4864) 
6.42, p  .05, but the associations varied among the
Ethnicity  Poverty Status groups, F(2, 4641)  3.78,
p  .05. Overall, spanking was modestly and nega-
tively associated with PPVT scores, but this associa-
tion was substantially more negative among the Afri-
can American nonpoor children,   .73, and less
negative among the African American poor children,
  .15, whereas it was substantially more negative
among the Hispanic American poor children,   .65,
and more positive among the Hispanic American non-
poor children,   .35.
The partial correlations between the HOME-SF
teaching factor and PPVT-R scores (controlling for
maternal education, father presence, and number of
household members) were generally not significant for
3- and 4-year-olds (even when significant they were
.20). Teaching, however, was significant for African
American and poor Hispanic American 5-year-olds
(see Table 7). Likewise, the partial correlations be-
tween the HOME-SF physical environment factor and
PPVT-R were generally not significant, albeit several
of the coefficients were marginal (about .10 to .15).
BPI
The HLM analyses revealed a complicated set of
relations between HOME-SF scores and children’s be-
havior problems (see Table 4). The block test for
HOME-SF factor scores was significant, F(3, 7103) 
83.2, p  .001. So, too, were the HOME-SF  Ethnic
Group Membership, F(6, 7155)  8.23, p  .001, and
HOME-SF  Poverty Group Membership  Ethnic
Group Membership, F(6, 6739)  2.15, p  .05 interac-
tions, and the Age  HOME-SF, F(3, 4793)  3.15, p 
.05, Age  HOME-SF  Ethnic Group Membership,
F(6, 4831)  2.90, p  .01, and Age  HOME-SF 
Ethnic Group Membership  Poverty Group Mem-
bership, F(4, 7863)  3.28, p  .05 interactions. Over-
all, fewer behavior problems were observed among
children from families with more positive HOME-SF
1876 Child Development

Table 4 HLM Results: Behavioral Problems Index

Regression Coefficients for Each Ethnicity  Poverty Status Group

F Value for Total Sample
European African Hispanic
American American American HOME 
HOME  HOME  Ethnicity 
Poor Nonpoor Poor Nonpoor Poor Nonpoor Overall  HOME Ethnicity Poverty Poverty

HOME-SF scales 83.2*** 8.23*** 1.85 2.15*

Learning 7.31*** 5.70*** 4.36*** 1.78*** .80 3.18 3.88*** 31.1*** 4.07* 1.21
stimulation (1.90) (1.05) (1.29) (1.41) (2.14) (1.69) (.69) a*
Parental .45 .90 3.32 .82 2.23 .24 .35 .49 3.63* 2.35
responsiveness (1.38) (.81) (1.01) (1.03) (1.58) (1.38) (.51) b*
Spanking 1.19*** 2.10*** .44*** 1.06*** 1.72*** 1.44*** 1.33*** 205.7*** 16.02*** 2.76
(.20) (.22) (.08) (.16) (.29) (.30) (.09) b***, a***
Age  HOME-SF scales 3.15* 2.90** 2.13 3.28*
Learning .30 .21 .32 .08 .98 .41 .06 .05 .36 .94
stimulation (.76) (.41) (.49) (.53) (.80) (.62) (.25)
Parental .06 .05 .27 .29 .21 .17 .20 .11 .33
responsiveness (.50) (.28) (.35) (.35) (.58) (.43)
Spanking .16** .31*** .15*** .06 .26*** .09 .09** 9.04** 8.48*** 5.82**
(.06) (.07) (.04) (.05) (.06) (.11) (.03) b**, a*, d* c**
Note: Values in parentheses represent standard errors. HLM  hierarchical linear modeling; HOME-SF  Home Observation for Measure-
ment of the Environment–Short Form.
a  a significant interaction involving African Americans versus European Americans.
b  a significant interaction involving African American versus Hispanic Americans.
c  a significant third-order interaction involving African Americans versus Hispanic Americans versus poverty versus nonpoverty.
d  a significant third-order interaction involving African Americans versus European Americans versus poverty versus nonpoverty.
* p  .05; ** p  .01; *** p  .001.

Table 5 HLM Results: Peabody Individual Achievement Test—Math

Regression Coefficients for Each Ethnicity  Poverty Status Group

F Value for Total Sample
European African Hispanic
American American American HOME 
HOME  HOME  Ethnicity 
Poor Nonpoor Poor Nonpoor Poor Nonpoor Overall  HOME Ethnicity Poverty Poverty

HOME-SF scales 17.2*** 1.61 1.62 .49

Learning 3.64 7.74*** 2.25 3.96** 3.22 2.94 3.98*** 33.0***
stimulation (1.91) (1.02) (1.23) (1.33) (2.18) (1.70) (.69)
Parental 1.98 .41 1.69 .42 3.77* 1.34 1.60** 10.7***
responsiveness (1.28) (.81) (.91) (.93) (1.57) (1.39) (.49)
Spanking .44* .18 .24** .36* .28 .17 .13 1.60
(.21) (.24) (.09) (.16) (.34) (.32) (.10)
Age  HOME-SF scales 3.06* .59 2.24 .99
Learning .30 .20 .53 .26 1.41 .38 .34 1.49
stimulation (.85) (.44) (.51) (.57) (.96) (.70) (.29)
Parental .70 .32 .76* .41 .89 .26 .47* 5.64*
responsiveness (.54) (.31) (.37) (.37) (.70) (.51)
Spanking .04 .03 .02 .02 .09 .22 .02 .52
(.08) (.09) (.04) (.05) (.08) (.13) (.03)

Note: Values in parentheses represent standard errors. HLM  hierarchical linear modeling; HOME-SF  Home Observation for Measure-
ment of the Environment–Short Form.
* p  .05; ** p  .01; *** p  .001.

Scale scores, especially when children were older. The
pattern of association varied as a function of both eth-
nicity and poverty status, however.
The HLM analyses showed that scores on the BPI
were negatively associated with scores on the learn-
ing stimulation factor of the HOME-SF, F(1, 7524) 
31.1, p  .001, but its association with BPI varied as a
function of ethnicity, F(2, 7566)  4.07, p  .05. Pair-
wise comparisons indicated that associations between
learning stimulation and the BPI were stronger among
European American children,   7.31 for poor chil-
dren and   5.70 for nonpoor children, than among
African American children,   4.36 for poor chil-
dren and   1.78 for nonpoor children.
The HLM analyses indicated that parental respon-
siveness was independently associated with children’s
behavior problems, but only for some subgroups of
children. The main effect for parental responsiveness
was nonsignificant, but the interaction between pa-
rental responsiveness and ethnicity was significant,
F(2, 6661)  3.63, p  .05. In contrast with learning
stimulation, pairwise comparisons suggested that pa-
rental responsiveness was associated with BPI scores
more negatively for African American children,  
3.32 for poor children and   .82 for nonpoor chil-
dren, than for European American children,   .45 for
poor children and   .90 for nonpoor children.
Finally, the HLM analyses showed a significant rela-
tion between the amount of spanking and BPI scores,
F(1, 6796)  205.7, p  .001, but this association was
different for the three ethnic groups, F(2, 6645)  16.02,
p  .001, across age groups, F(1, 4544)  9.04, p  .01,
and among different ethnic and poverty status groups
across ages, F(2, 4706)  5.82, p  .01. As indicated by
these interactions, there were very complex interac-
tions among amount of spanking, child age, ethnicity,
and poverty status. Spanking showed a stronger rela-
tion with the BPI in Hispanic Americans when chil-
dren were younger, but a stronger relation with the
BPI in African American when children were older.
For example, the change in BPI scores across time as-
sociated with reporting one more spanking per week
was .26 (Age  Spanking coefficient) for poor His-
panic American children, .15 for poor African Ameri-
can children, and .16 for poor European American
children. This translates into a much stronger nega-
tive effect at younger ages for the Hispanic American
children and a stronger negative effect at older ages
for the African American and European American
children. The differences in age trends between Afri-
can American and Hispanic American families were
stronger for poor than nonpoor families, (that is,  
.26 and .15 for poor Hispanic American and African
American children, respectively, versus   .06 and
Bradley et al. 1877
.09 for nonpoor African American and Hispanic
American children, respectively. The differences in
age trends between African American and European
American families were stronger for nonpoor than for
poor families.
There was scattered evidence that children’s be-
havior problems were associated with parental teach-
ing; the few significant relations were in the expected
negative direction (see Table 7). There was little evi-
dence that the physical environment factor was re-
lated to BPI scores; only 2 of 24 coefficients were sta-
tistically significant.
PIAT Math
In the HLM analyses performed on PIAT math
scores, the block test for HOME-SF factor scales was
significant, F(3, 6053)  17.2, p  .001, but the magni-
tude of their association varied with age, F(3, 3738) 
3.06, p  .05. No reliable differences in association be-
tween the three HOME-SF scales and PIAT math
scores pertaining to ethnic group membership or pov-
erty status membership were detected (see Table 5).
The HLM analyses showed a significant relation be-
tween PIAT math scores and both learning stimula-
tion, overall   3.98, F(1, 6405)  33.0, p  .001, and
parental responsiveness, overall   1.60, F(1, 5704) 
10.7, p  .001, controlling for all the sociodemographic
and household composition measures in the model.
There were no interactions between learning stimula-
tion and age, ethnicity, or poverty status, suggesting
that all children of all ages tended to show higher math
scores if the children were provided more learning
stimulation. In contrast, the Age  Parental Respon-
siveness interaction, overall   .47, F(1, 4374) 
5.64, p  .05, indicated that parental responsiveness
was more strongly associated with higher math scores
among younger children than among older children.
For example, the expected difference in math scores
for children whose homes received credit for none of
the parental responsiveness items versus children from
homes that received credit for all the items was 1.60 at
the mean age of the sample (due to the fact that all vari-
ables were centered), but was 2.54 two years earlier,
1.60 (2  .47  2.54). The expected difference in
math scores was .19 at 3 years older than the mean age
of the sample, 1.60 (3  .47  .19). Spanking was not
significantly related to math scores in these analyses.
Somewhat surprisingly, there was little evidence
that math performance was associated with parental
teaching (see Table 7). The only significant partial cor-
relation between parental teaching and PIAT math
was among nonpoor African Americans, r(222)  .24.
Nor was there much evidence that math performance
1878 Child Development

Table 6 HLM Results: Peabody Individual Achievement Test—Reading Recognition

Regression Coefficients for Each Ethnicity  Poverty Status Group

F Value for Total Sample
European African Hispanic
American American American HOME 
HOME  HOME  Ethnicity 
Poor Nonpoor Poor Nonpoor Poor Nonpoor Overall  HOME Ethnicity Poverty Poverty

HOME-SF scales 18.9*** 1.32 .58 2.51*

Learning 2.76 7.75*** 3.44** 3.00* 2.59 2.90 3.76*** 2.12
29.2***
stimulation (1.90) (1.04) (1.22) (1.32) (2.20) (1.70) (.69)
Parental .43 1.60* 1.39 .81 4.53** 2.47 1.72** 13.3*** 1.60
responsiveness (1.24) (.77) (.87) (.89) (1.55) (1.34) (.47)
Spanking .56** .07 .09 .53** .28 .08 .27** 7.13** 3.74*
(.21) (.24) (.09) (.16) (.33) (.33) (.10) a**
Age  HOME-SF scales 2.55 1.60 1.27 .77
Learning 1.69* .03 .49 .54 1.11 .44 .23
stimulation (.85) (.45) (.51) (.57) (.97) (.70) (.29)
Parental .18 .46 .38 .55 .52 .67 .46*
responsiveness (.54) (.31) (.37) (.36) (.71) (.50) (.20)
Spanking .11 .16 .01 .07 .03 .12 .03
(.08) (.09) (.04) (.05) (.08) (.14) (.03)

Note: Values in parentheses represent standard errors. HLM  hierarchical linear modeling; HOME-SF  Home Observation for Measure-
ment of the Environment–Short Form.
a  a significant third-order interaction involving African Americans versus European Americans versus poverty versus nonpoverty.
* p  .05; ** p  .01; *** p  .001.

Table 7 Partial Correlations between HOME-SF Teaching and Physical Environment Scales and Child Outcomes at Various Ages

European American African American Hispanic American

Age Poor Nonpoor Poor Nonpoor Poor Nonpoor

HOME-SF teaching
PPVT-R 3 and 4 .12 .14** .00 .07 .16 .11
5 .05 .10 .20* .28** .29 .11
BPI 4 .22** .00 .02 .27** .15 .04
5 .01 .11 .04 .02 .32 .00
PIAT math 5 .09 .06 .04 .24** .02 .10
PIAT reading 5 .16 .05 .20* .22** .27 .03
HOME-SF physical environment
PPVT-R 3 and 4 .11 .13** .14 .14 .09 .01
6 through 9 .15** .01 .01 .02 .07 .00
BPI 4 .05 .03 .11 .11 .02 .06
5 .14 .08 .11 .02 .13 .15
6 and 7 .08 .08 .11 .05 .09 .04
8 and 9 .18** .08* .01 .01 .15 .03
PIAT math 5 .21 .04 .00 .06 .10 .22*
6 and 7 .11 .05 .09 .00 .00 .11
8 and 9 .12 .03 .04 .03 .01 .00
PIAT reading 5 .22 .09 .08 .05 .20 .01
6 and 7 .11 .08 .04 .02 .14 .04
8 and 9 .05 .01 .05 .00 .08 .00

Note: All coefficients are partial correlations controlling for father presence, number of household members, and maternal education.
HOME-SF  Home Observation of the Environment–Short Form; PPVT-R  Peabody Picture Vocabulary Test–Revised; BPI  Behavior
Problems Index; PIAT  Peabody Individual Achievement Test.
* p  .05; ** p  .01.

was associated with the quality of the physical envi-
ronment. The only group in which significant rela-
tions were observed was nonpoor Hispanic Ameri-
cans, r(196)  .22.
PIAT Reading Recognition
For PIAT reading recognition, the overall block test
for the HOME-SF factor scales was significant,
F(3, 5580)  18.9, p  .001. As Table 6 shows, however,
the interaction among HOME-SF, poverty, and eth-
nicity was also significant, F(6, 5131)  2.51, p  .05.
Neither the interaction between HOME-SF and pov-
erty nor the interaction between HOME-SF and eth-
nicity was significant.
The HLM analyses showed significant relations be-
tween PIAT reading recognition scores and all three
HOME-SF factors: learning stimulation overall  
3.76, F(1, 5985)  29.2, p  .001; paternal responsive-
ness, overall   1.72, F(1, 5065)  13.3, p  .001; and
spanking, overall   0.27, F(1, 5497)  7.13, p  .01.
The relation between spanking and reading was com-
plex. The relation was stronger for European American
poor children,   .56, than for African American poor
children,   .09. In contrast, the association was
stronger for African American nonpoor children,  
.53, than for European American nonpoor children,
  .07. Although there were isolated instances of
significant Age  HOME-SF effects, for example,  
1.69 for poor European Americans, the block test for
the Age  HOME-SF interaction was not significant.
For 5-year-old children, parental teaching was re-
lated to PIAT reading among African Americans. There
were no significant partial correlations between phys-
ical environment scores and PIAT reading recognition
scores, however; two coefficients approached signifi-
cance, but both were low.
DISCUSSION
The purpose of Part 1 of this study of NLSY children
and families (Bradley et al., 2001) was to describe the
experiences that children in the United States have in
their home environments. The purpose of Part 2 (the
present study) was to describe the relation between
those experiences and major aspects of the children’s
development from birth through age 13: early motor
and social development, language development (re-
ceptive vocabulary), achievement, and behavior prob-
lems. To that end we investigated relations between
these aspects of development and three classes of en-
vironmental action that have been the subject of nu-
merous developmental studies over the past 40 years;
Bradley et al. 1879
namely, materials and experiences that stimulate learn-
ing, parental responsiveness, and spanking. For more
restricted age ranges we were also able to examine re-
lations for the physical environment and teaching as
well. As with the analyses performed in Part 1, we stud-
ied these relations for the three largest ethnic groups in
America (European Americans, and African Ameri-
cans, Hispanic Americans), each broken into two in-
come groups (poverty and nonpoverty).
Prior to discussing the findings, it is important to
state that the sample of indicators used to represent each
home environment domain is nowhere near the full cen-
sus of indicators for that domain. For that reason, esti-
mates of relations between these five domains and the
four aspects of development examined are likely to be
somewhat attenuated (see Bollen & Lennox, 1991). It is
also important to restate that we have used the primary
ethnic group designations established by the NLSY in
analyzing the data. Although these designations are
useful from the standpoint of connecting findings to the
broad literature on child development, they, nonethe-
less, mask substantial within-group variation for each of
the major ethnic groups studied (Gjerde & Onishi, 2000).
Learning Stimulation
Of the major dimensions of the environment stud-
ied, the most consistent set of findings were those in-
volving learning stimulation. Learning stimulation was
associated with early motor and social development,
language competence, and achievement in all three
ethnic groups and at almost every age for both poor
and nonpoor children. Results showing small, but sig-
nificant, relations between learning stimulation and
motor and social development in the first 3 years of life
are consistent with findings from previous studies con-
ducted with smaller samples, consisting mostly of Eu-
ropean Americans and African Americans (Bradley,
1994). The amount of variance accounted for during
the toddler period was somewhat less than has been
typically observed in studies that used the full Bayley
scales or other comparable measures of early develop-
ment (Bradley, 1994; McCall, 1981). These findings
may reflect the fact that the NLSY motor and social de-
velopment measure was relatively brief and tapped
mostly motor and self-care routines. These two areas of
development are not as likely as cognitive and lan-
guage development (areas also assessed by the full Bay-
ley Scales) to be affected by the kinds of indicators in-
cluded in the HOME-SF learning stimulation item set.
Learning stimulation was also moderately associ-
ated with language competence, as measured by the
PPVT-R. The HLM analyses indicated that the strength
of the relation was similar for poor and nonpoor chil-
1880 Child Development
dren and for children from all three ethnic groups. Al-
though the overall block test for HOME-SF scales in-
dicated no age differences in the strength of relations
between the home environment and vocabulary per-
formance, there was a significant interaction between
learning stimulation and age. The results indicate that
the effect of learning stimulation is slightly stronger
for younger children. Taken together, studies on the
relation between learning stimulation and language
competence indicate a robust relation (Broberg, Wes-
sels, Lamb, & Hwang, 1997; Jordan, 1978; Luster &
Dubow, 1992). In the present study, learning stimulation
was associated with size of vocabulary, even when con-
trolling for sociodemographic factors and other inputs
from the family environment, such as parental respon-
siveness. In the Luster and Dubow study, HOME-SF to-
tal score was correlated with vocabulary during the pri-
mary grades, even when controlling for maternal IQ. In
the Colorado Adoption Study, Plomin and colleagues
(Braungart, Fulker, & Plomin, 1992; Plomin, Loehlin, &
DeFries, 1985) found little evidence of genetic mediation
regarding child competence for items on the full HOME
Inventory that tap learning stimulation.
As expected, learning stimulation was also signifi-
cantly correlated with achievement test scores among
kindergarten-age children (Bradley & Caldwell, 1995;
Cairns, Cairns, Xie, Leung, & Hearne, 1998; Kagan,
1984). The HLM analyses indicated no interactions with
ethnicity and provided little evidence that the strength
of the relation between learning stimulation and
achievement test scores changes from the primary
grades to middle school. Gottfried, Fleming, and Gott-
fried (1998) presented recent longitudinal data from
children (ages 8–13) living in Fullerton, California that
showed a relation between cognitively stimulating as-
pects of the home environment and children’s academic
intrinsic motivation. They argued that the impact of the
home environment on academic intrinsic motivation
mediates the relation with academic performance.
What is a bit more surprising is the consistent asso-
ciation between learning stimulation and behavior
problems. Most of the literature that addresses the en-
vironmental correlates of children’s behavior problems
focuses on the social aspects of family environment
(discipline strategies, warmth, responsiveness, and the
like; Conger et al., 1992; Deater-Decker et al., 1996; Flor-
sheim et al., 1998; McLoyd, 1998; Neiderhiser, Reiss,
Hetherington, & Plomin, 1999; Patterson, 1982; Pettit
et al., 1997). Relatively little attention has been focused
on material aspects of the environment or efforts by
parents to provide enriching learning experiences.
The fact that learning stimulation in this study was
associated with children’s behavior problems in ev-
ery ethnic group from early childhood through ado-
lescence, despite controlling for family demograph-
ics, parental responsiveness, and level of spanking,
suggests that theories about the causes and mainte-
nance of behavior problems may need to be expanded
to consider the roles played by the object environ-
ment and direct efforts of parents to provide learning
opportunities for their children.
Teaching
The findings pertaining to maternal teaching—
although restricted to the preschool period—are remi-
niscent of the findings for learning stimulation, albeit
weaker. Partial correlations between maternal teach-
ing and children’s competence were quite low, rarely
exceeding .20. It is not clear why the observed associa-
tions between maternal teaching and children’s com-
petence were so marginal, given the theoretical linkage
between the two. Part of the reason almost certainly
pertains to the fact that we controlled for maternal ed-
ucation. Perhaps more than any other HOME-SF fac-
tor, the amount that mothers teach their children re-
flects their own level of education. The small residual
correlations may also reflect limitations in the HOME-
SF items used to assess maternal teaching practices.
The items only distinguished between those mothers
who taught certain concepts and those who did not.
The items made no effort to determine how frequently
or how effectively the concepts were taught.
Parental Responsiveness
As expected, parental responsiveness showed a rela-
tion to early motor and social competence. These find-
ings, although small in magnitude, are consistent with
the generally acknowledged value of social responsive-
ness as supportive of secure attachment and the longer
term value such early responsiveness may have on
social competence (Bretherton, 1985). Theory also stipu-
lates that responsiveness continues to play a role in
adaptive behavior for older children (Bogenschneider,
Wu, Raffaelli, & Tsay, 1998; Brody & Flor, 1998; McLoyd,
1999). This study provided limited support for such a
proposition in that the main effect for paternal re-
sponsiveness for children’s behavior problems was
nonsignificant. Results from the HLM analyses, how-
ever, indicated that among African Americans (espe-
cially poor), more parental responsiveness was asso-
ciated with fewer behavior problems (see also Deater-
Decker et al., 1996). In contrast, there was a tendency
(not statistically significant) among Hispanic Ameri-
cans (especially nonpoor) for more parental respon-
siveness to be associated with more behavior prob-
lems. The reason for these small ethnic and poverty

status differences is not fully evident. The low posi-
tive association for Hispanic Americans may partly
reflect the fact that many Hispanic families are
strongly connected to a network of kin in which care-
giving is more distributed than may be the case in
other ethnic groups (Gamble & Dalla, 1998). When
children interact with a number of different care-
givers and family members, the influence of any one (in
this case, mothers) may be somewhat lessened. Relat-
edly, the low association may partly reflect a sense of
familism, which carries with it a press to maintain
positive relations with other family members and
some degree of reticence to perceive family members in
a negative light (Cooper, 1999; Gamble & Dalla, 1998).
The failure to find strong relations between paren-
tal responsiveness and the BPI was somewhat at vari-
ance with results from a study reported by Shaw,
Keenan, and Vondra (1994). In their study, parental
responsiveness, measured at 12 months, showed a
low, but significant correlation with 3-year externaliz-
ing behaviors. Likewise, Wakslag and Hans (1999)
found that early parental responsiveness was related
to later behavior problems. Several aspects of these
other two studies were different, including the fact
that neither controlled for family demographics or
other aspects of the home environment. Further in-
vestigation of this particular relation is warranted.
Responsiveness is thought to play a role in the
early development of competence (Bornstein, 1989;
Pettit et al., 1997). Results from this study showing a
relation between parental responsiveness and perfor-
mance on both the PPVT-R and the PIAT provide sup-
port for this proposition. The significant interaction
between parental responsiveness and age for PIAT
math (the trends were also evident for PIAT reading
and PPVT-R, but neither was statistically significant)
indicates that the relation is stronger when children
are young. Overall, the HLM findings suggest that as
children age, achievement becomes increasingly con-
nected to family efforts to promote achievement.
There is also evidence that achievement is connected to
the demands and supports for achievement in other
contexts, such as school and peer groups (Brown, 1999;
Scott-Jones, 1995). In sum, the relation (and presumed
effect) of parental responsiveness and children’s com-
petence weakened as children got older. The inconsis-
tent age interaction observed in this study provides
only partial support for Scarr’s (1992) argument that
the impact of home environment decreases as chil-
dren mature and develop the skills to seek out and
fashion environments that are suited to their inter-
ests. Perhaps it is more accurate to say, as Wachs
(2000) suggests, that these relations appear to become
more complex as children get older. It is not, for ex-
Bradley et al. 1881
ample, that parental responsiveness is unimportant
for later development, but its influence becomes more
dependent on other aspects of family context (Bradley
& Corwyn, 2000), other ecological factors (Fletcher,
Darling, Steinberg, & Dornbusch, 1995; Harkness &
Super, 1995), and the position of the family vis-à-vis
the larger society (García Coll & Magnusson, 1999).
Spanking
Perhaps the most complex set of relations observed
between HOME-SF factors and measures of child de-
velopment were those involving spanking. A relation
between spanking and behavior problems was ob-
served in all groups. However, there were quite com-
plex interactions between spanking and behavior
problems as a function of ethnicity and poverty sta-
tus. For example, the relation between spanking and
behavior problems became stronger as children aged
for European Americans and poor African Ameri-
cans, whereas the trend was reversed for poor His-
panic Americans.
The findings pertaining to spanking were gener-
ally consistent with previous research and theory
(Day, Peterson, & McCracken, 1998; Florsheim et al.,
1998; Pettit et al., 1997). For example, as expected,
amount of spanking was related to the number of be-
havior problems observed in adolescence (Goodman
et al., 1998; Vuchinich, Bank, & Patterson, 1992). As
stated earlier, for every group except poor Hispanic
Americans, the correlations were stronger as children
grew older. Prior to school entry, there was no relation
between the use of harsh punishment and parental
perceptions of child behavior problems in any of the
groups examined. During the elementary school years,
there were low correlations in all three ethnic groups,
although the relations were less consistent among Af-
rican Americans. By the time children reached second-
ary schools, however, the relations were quite consis-
tent and substantially higher, even as the actual use of
punishment decreased substantially. The analyses did
not permit inferences regarding direction of effect but
they were consistent with the proposition that acting
out on the part of older children evokes negative re-
sponses on the part of parents (Neiderhiser et al., 1999).
During the transition to adolescence, it is common for
both parents and children to alter their perception of
the relationship somewhat (Collins et al., 1995; Holm-
beck et al., 1995), each tending to attribute more nega-
tive qualities to the motivations and behavior of the
other. The stronger association between spanking and
BPI scores may be reflective of the parent’s perception
of the adolescent as defiant and uncontrollable, with a
corresponding attempt to exercise power-assertive
1882 Child Development
control. These difficulties may lead to marital conflict,
which may increase parental hostility toward the ad-
olescent and subsequently lead to adolescent distress
and acting out (Harold & Conger, 1997). On the other
hand, the increased correlation between the use of
spanking and BPI scores could reflect a cumulative
history of harsh punishment as well, with some par-
ents continuing to mete out harsh punishment that
has the effect of exacerbating acting out behaviors
(Patterson, 1982). Other researchers have also observed
that the relation between harsh parenting and behav-
ior problems increases with age (Pettit et al., 1997).
Overall, the relatively weaker association observed
between spanking and behavior problems for African
Americans fits with the rather inconsistent pattern of
results for African American children obtained in other
studies; some studies show a positive relation, some a
negative relation, and some no relation (Whaley, 2000).
It may be that physical punishment, short of abuse, has
a different meaning in African American families, in
which there has been a history of oppression (Deater-
Decker et al., 1996; McLoyd et al., 2000). Especially for
poor African Americans, punishment may afford some
protection against the adverse consequences of violat-
ing social rules (Kelly, Power, & Winbush, 1992). The
results for Hispanic Americans are consonant with re-
sults reported by Lindahl and Malik (1999) indicating
that in Latino families strictness is not associated with
externalizing behaviors. Unlike European American
families, it appears that punishment in Hispanic Amer-
ican families is not associated with a cold and insensi-
tive style of parenting.
Spanking was unrelated to early motor and social
development. Its relation to later competence tended
to be weak and somewhat inconsistent across groups.
For both PPVT-R and PIAT reading, there was a signif-
icant three-way interaction among spanking, ethnicity,
and poverty status. The strongest relations observed
were for nonpoor African Americans and for poor Eu-
ropean Americans. These findings suggest (but do not
demonstrate) that, for some groups, spanking may be a
response to poor academic performance. Day and col-
leagues (1998) found that mothers spanked more when
they perceived their children to be less competent.
However, Pettit and colleagues (1997) found that harsh
parenting during kindergarten was correlated equally
strongly with academic performance during kinder-
garten and sixth grade, so the relation may be quite
complex with influences flowing in both directions.
Physical Environment
Unfortunately, the fact that the physical environ-
ment was only measured when children were be-
tween the ages of 3 and 9 makes it difficult to draw
strong conclusions regarding its relation to children’s
development. As expected, children living in higher
quality physical environments showed higher levels
of competence and lower levels of behavior prob-
lems. These associations were observed somewhat
more often in poor families. The relatively small ef-
fects observed in this study may be somewhat decep-
tive. Relations between the physical environment and
measures of children’s well-being were calculated
within economic subgroups (thereby restricting the
range), controlling for other demographic factors
known to be associated with the quality of the physi-
cal environment.
Consistent with previous research, the proximal as-
pects of the environment assessed by HOME-SF were
significantly associated with achievement even when
controlling for family income, maternal education,
family size, ethnicity, and father presence (for reviews,
see Bradley, 1994 and Wachs, 1992). As Bloom (1964)
argued nearly 40 years ago and as we showed in Part
1 of this study (Bradley et al., 2001), there is variability
in what children experience in their home environ-
ments in every demographic group. Furthermore,
variability in those experiences connects to differences
in children’s developmental trajectories. That said,
these analyses do not allow for making causal infer-
ences regarding the relations observed and they do
not address the issue of genetic mediation.
This study provided only limited support for the
idea that relations between proximal home factors
and achievement test performance diminish with age.
Occasionally there were statistically significant age
effects, but the findings were inconsistent across mea-
sures and groups. The approach to data analysis used
in this study may not have been ideal from the stand-
point of understanding whether processes connected
with other environments (e.g., schools, peer groups,
neighborhoods) gradually assume control of achieve-
ment behaviors. Controlling for key family demo-
graphic factors in the analyses may have inadvert-
ently resulted in controlling for key outside-of-family
factors as well. At present hypotheses about such
“turn over” effects remain largely undocumented (for
an extended discussion of the school effects research,
see Talbert & McLaughlin, 1999).
The findings pertaining to behavior problems sug-
gest that there is no simple relation between what
children experience at home and the likelihood they
will manifest behavioral difficulties. All five aspects
of the home environment studied showed some asso-
ciation with behavior problems in children, but rela-
tions varied by ethnic group, poverty status, and age.
The findings are consistent with the broader literature

on child adjustment, which indicates that predicting
adjustment in individual children is very difficult ex-
cept at the extremes. There are currently a number of
very useful models linking various aspects of parent-
ing (e.g., harsh punishment) to maladaptive behavior
in children (e.g., Brody & Flor, 1998; Conger et al., 1992;
McLoyd, 1998). Findings from this study indicate that
several parenting factors may be implicated in mal-
adaptive behavior and that models linking parenting
to child behavior problems may need to be expanded
to include not only direct actions by parents but phys-
ical aspects of the home environment as well.
Although the primary purpose of this study was
to examine relations within ethnic groups, we tested
several interactions with ethnicity as part of the HLM
analyses. Few significant interactions between eth-
nicity and home environmental processes emerged.
When they did, the associations tended to be some-
what stronger for European Americans (especially
nonpoor). Such findings are consistent with the idea
that the kinds of indicators found on commonly used
measures of the home environment are somewhat
more culturally relevant for middle-class European
Americans than for people of color (Bradley et al.,
2000). It is also consistent with the idea that there is
sometimes a clash between the developmental scripts
used by minority group parents and the scripts of the
predominant culture in the United States. Even so,
higher order interactions were less frequently ob-
served and less consistent and there were more simi-
larities across ethnic groups than dissimilarities.
Findings from Part 1 of this study (Bradley et al.,
2001) made clear that what children experience in
their home environments varies greatly both within
and across the major income and ethnic groups in the
United States. There are also characteristic shifts in
the likelihood a child will encounter certain types of
experience from infancy through adolescence (e.g.,
the amount of spanking). Accordingly, there tends to
be only moderate stability in most dimensions of the
environment across time—albeit the small numbers
of indicators of most environmental dimensions in
the NLSY may have led to some underestimates of
stability in these dimensions. The findings from Part 2
indicate that both children’s competence and their be-
havior are associated with proximal aspects of their
home environments. Learning stimulation, in partic-
ular, was associated with every aspect of develop-
ment examined in nearly every group at nearly every
age, even when controlling for family demographics
and other aspects of the social environment. In con-
trast, findings from Part 2 indicated that most of the
variance in children’s competence and behavior is not
attributable to the kinds of experiences contained in
Bradley et al. 1883
commonly used home environment measures (such as
is represented by the HOME short forms). In general,
no one dimension accounted for more than 10% of the
variance, nor did the combination of dimensions ac-
count for more than 30%. Moreover, the amount of
variance accounted for differed by ethnicity and pov-
erty status. In overview, however, the findings per-
taining to ethnicity were more remarkable for their
similarities than their differences. If the HOME-SF had
contained a more complete census of indicators for
each of the home environment dimensions indexed
and if we had not controlled for family demography
in the HLM analyses, perhaps more ethnic differences
would have emerged. In essence, there is a need to ex-
amine these relations using a more complete census
of indicators for the kinds of major environmental di-
mensions included in this study and to study the re-
lations on more targeted ethnic groups to more au-
thoritatively determine the degree of universality of
such relations (Bornstein, 1995a).
Wachs (2000) has argued that there is enormous
variability in developmental outcomes for children
who have similar experiences at home (or in other con-
texts). He stated that “individual variability is a neces-
sary consequence of complex interactions among
multiple influences that are each necessary but not
sufficient contributors to behavioral development”
(p. 3). Our analyses of the NLSY data, although not
structured to detect parental influences on develop-
ment, support the idea that parental influence is not
simple and straightforward. Rather it appears to be
highly complex and differentiated (for a discussion of
the parenting literature, see Collins, Maccoby, Stein-
berg, Hetherington, & Bornstein, 2000).
Finally, the findings attest to the value of examin-
ing relations between children’s development and
several aspects of children’s environments simulta-
neously. Children do not experience each aspect of
their environments in isolation from the others. There
is increasing evidence of modularity, additivity, and
interactivity in the relations between specific aspects
of children’s environments and developmental course
in most developmental domains (Bornstein & Tamis-
LeMonda, 1997; Bradley & Caldwell, 1995; Wachs,
1992). Understanding the relation of each aspect of the
home environment to patterns of development will be
made clearer when simultaneously considering other
aspects. The idea of studying multiple proximal fac-
tors of children’s home environments simultaneously
applies to other proximal ecologies as well. Children
have experiences in multiple ecologies throughout
childhood (child care as well as home, school as well
as home, peer group as well as school). Studying si-
multaneous proximal inputs from each in relation to
1884 Child Development
the others may enable us to better understand how
environments operate to influence the course of de-
velopment from infancy through adolescence.
ACKNOWLEDGMENTS
Funding for this research was received from the Office
of Educational Research and Improvement through a
grant to the National Center for Early Learning and
Development. The authors wish to thank Frank Mott
and Paula Baker for their assistance with the NLSY
dataset. More detailed information on findings from
this study can be obtained at http://www.ualr.edu/
~HOME.
ADDRESSES AND AFFILIATIONS
Corresponding author: Robert H. Bradley, Center for
Applied Studies in Education, University of Arkansas
at Little Rock, 2801 S. University Avenue, Little Rock,
AR 72204; e-mail: rhbradley@ualr.edu. Robert F. Cor-
wyn is also at University of Arkansas; Margaret
Burchinal is at the University of North Carolina at
Chapel Hill; Harriette Pipes McAdoo is at Michigan
State University, East Lansing, MI; and Cynthia García
Coll is at Brown University, Providence, RI.
REFERENCES
Baker, P. C., Keck, C. K., Mott, F. L., & Quinlan, S. V. (1993).
NLSY child handbook–revised edition: A guide to the 1986–
1990 National Longitudinal Survey of Youth child data. Co-
lumbus: Ohio State University, Center for Human Re-
source Research.
Bloom, B. (1964). Stability and change in human characteristics.
New York: Wiley.
Bogenschneider, K., Wu, M., Raffaelli, M., & Tsay, J. C.
(1998). Parent influences on adolescent peer orientation
and substance use: The interface of parenting practices
and values. Child Development, 69, 1672–1688.
Bolger, K. E., Patterson, C. J., Thompson, W. W., & Ku-
persmidt, J. B. (1995). Psychological adjustment among
children experiencing persistent and intermittent family
economic hardship. Child Development, 66, 1107–1129.
Bollen, K. A., & Lennox, R. (1991). Conventional wisdom on
measurement: A structural equation perspective. Psycho-
logical Bulletin, 110, 305–314.
Bornstein, M. H. (1989). Maternal responsiveness: Characteris-
tics and consequences. San Francisco, CA: Jossey-Bass.
Bornstein, M. H. (1995a). Form and function: Implications
for studies of culture and human development. Culture
and Psychology, 1, 123–137.
Bornstein, M. H. (1995b). Parenting infants. In M. H. Bornstein
(Ed)., Handbook of parenting: Vol. 1. Children and parenting
(pp. 3–40). Mahwah, NJ: Erlbaum.
Bornstein, M. H., & Tamis-LeMonda, C. S. (1997). Maternal
responsiveness and infant mental abilities: Specific pre-
dictive relations. Infant Behavior and Development, 20,
283–296.
Bradley, R. H. (1994). The HOME Inventory: Review and re-
flections. In H. Reese (Ed.), Advances in child development
and behavior (Vol. 25, pp. 241–288). Orlando, FL: Aca-
demic Press.
Bradley, R. H., & Caldwell, B. M. (1995). Caregiving and the
regulation of child growth and development: Describing
proximal aspects of caregiving systems. Developmental
Review, 15, 38–85.
Bradley, R. H., Caldwell, B. M., Rock, S. L., Barnard, K.,
Gray, C., Hammond, M., Mitchell, S., Siegel, L., Ramey,
C., Gottfried, A., & Johnson, D. (1989). Home environ-
ment and cognitive development in the first 3 years of
life: A collaborative study involving six sites and three
ethnic groups in North America. Developmental Psychol-
ogy, 25, 217–235.
Bradley, R. H., & Corwyn, R. (2000). The moderating effect of
perceived family conflict on the relation between home
environmental processes and the well-being of adoles-
cents. Journal of Family Psychology, 14, 349–364.
Bradley, R. H., Corwyn, R. F., Caldwell, B. M., Whiteside-
Mansell, L., Wasserman, G. A., Walker, T. B., & Mink, I. T.
(2000). Measuring the home environments of children in
early adolescence. Journal of Research on Adolescence, 10,
247–288.
Bradley, R. H., Corwyn, R. F., McAdoo, H. P., & García Coll,
C. (2001). The home environments of children in the
United States Part I: Variations by age, ethnicity, and
poverty status. Child Development, 72, 1844–1867.
Bradley, R. H., & Tedesco, L. (1982). Environmental corre-
lates of mental retardation. In J. Lachenmeyer & M. Gibbs
(Eds.), The psychology of the abnormal child (pp. 155–189).
New York: Gardner.
Braungart, J. M., Fulker, D. W., & Plomin, R. (1992). Genetic
mediation of home environment during infancy: A sib-
ling adoption study of HOME. Developmental Psychology,
28, 1048–1055.
Bretherton, I. (1985). Attachment theory: Retrospect and
prospect. Monographs of the Society for Research in Child
Development, 50(1–2, Serial No. 209).
Broberg, A. G., Wessels, H., Lamb, M. E., & Hwang, C. P.
(1997). Effects of day care on the development of cogni-
tive abilities in 8-year-olds: A longitudinal study. Devel-
opmental Psychology, 33, 62–69.
Brody, G. H., & Flor, D. L. (1998). Maternal resources,
parenting practices, and child competence in rural, single-
parent African American families. Child Development, 69,
803–816.
Bronfenbrenner, U. (1995). The bioecological model from a
life course perspective: Reflections of a participant ob-
server. In P. Moen, G. H. Elder, & K. Luscher (Eds.), Ex-
amining lives in context (pp. 599–618). Washington, DC:
American Psychological Association.
Brown, B. B. (1999). Measuring the peer environment of
American adolescents. In S. L. Friedman & T. D. Wachs
(Eds.), Measuring environments across the life-span (pp.

91 – 126). Washington, DC: American Psychological
Association.
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear
models: Applications and data analysis methods. Newbury
Park, CA: Sage.
Cairns, R. B., Cairns, B. D., Xie, H., Leung, M., & Hearne, S.
(1998). Paths across generations: Academic competence
and aggressive behaviors in young mothers and their
children. Developmental Psychology, 34, 1162–1174.
Caldwell, B., & Bradley, R. H. (1984). Home Observation for
Measurement of the Environment. Little Rock: University
of Arkansas at Little Rock.
Center for Human Resource Research (CHRR). (1995).
NLS user’s guide 1995. Columbus: The Ohio State
University.
Center for Human Resource Research (CHRR). (1997a).
Mainfile codebook supplement, No. 2: Total net family income
variable creation 1979–1994. Columbus: Ohio State
University.
Center for Human Resource Research (CHRR). (1997b).
1994 child & young adult data users guide. Columbus: The
Ohio State University.
Chao, R. K. (1994). Beyond parental control and authoritar-
ian parenting style: Understanding Chinese parenting
through the cultural notion of training. Child Develop-
ment, 65, 1111–1119.
Collins, W. A., Harris, M. L., & Susman, A. (1995). Parenting
during middle childhood. In M. H. Bornstein (Ed.), Hand-
book of parenting: Vol. 1. Children and parenting (pp. 65–90).
Mahwah, NJ: Erlbaum.
Collins, W. A., Maccoby, E. E., Steinberg, L., Hetherington,
E. M., & Bornstein, M. H. (2000). Contemporary research
on parenting. American Psychologist, 55, 218–232.
Conger, R., Conger, K., Elder, G., Lorenz, F., Simolns, R., &
Whitbeck, L. (1992). A family process model of economic
hardship and adjustment of early adolescent boys. Child
Development, 63, 526–541.
Cooper, C. R. (1999). Multiple selves, multiple worlds: Cul-
tural perspectives on individuality and connectedness in
adolescent development. In A. S. Masten (Ed.), The Min-
nesota symposia on child psychology: Vol. 29. Cultural pro-
cesses in child development (pp. 25–57). Mahwah, NJ:
Erlbaum.
Day, R. D., Peterson, G. W., & McCracken, C. (1998). Predict-
ing spanking of younger and older children by mothers
and fathers. Journal of Marriage and Family, 60, 79–94.
Deater-Decker, K., Dodge, K. A., Bates, J. E., & Pettit, G. S.
(1996). Physical discipline among African American and
European American mothers: Links to children’s external-
izing behaviors. Developmental Psychology, 32, 1065–1072.
Duncan, G. P., Brooks-Gunn, J., & Klebanov, P. K. (1994).
Economic deprivation and early childhood develop-
ment. Child Development, 65, 296–318.
Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabu-
lary Test–Revised. Circle Pines, MN: American Guidance
Service.
Dunn, L. M., & Markwardt, F. C. (1970). Peabody Individual
Achievement Test manual. Circle Pines, MN: American
Guidance Services.
Bradley et al. 1885
Edwards, C. P. (1995). Parenting toddlers. In M. H. Born-
stein (Ed)., Handbook of parenting: Vol. 1. Children and
parenting (pp. 41–64). Mahwah, NJ: Erlbaum.
Fletcher, A., Darling, N., Steinberg, L., & Dornbusch, S.
(1995). The company they keep: Relation of adolescents’
adjustment and behavior to their friends’ perceptions of
authoritative parenting in the social network. Develop-
mental Psychology, 31, 300–310.
Florsheim, P., Tolan, P., & Gorman-Smith, D. (1998). Family
relationships, parenting practices, the availability of
male family members, and the behavior of inner-city
boys in single-mother and two-parent families. Child De-
velopment, 69, 1437–1447.
Ford, D. M., & Lerner, R. M. (1992). Developmental systems
theory. Newbury Park, CA: Sage.
Gamble, W. C., & Dalla, R. L. (1998). Parenting and child ad-
justment in single- and two-parent Euro- and Mexican-
American families. In H. I. McCubbin, E. A. Thompson,
A. I. Thompson, & J. E. Fromer (Eds.), Resiliency in Native
American and immigrant families (pp. 343–365). Thousand
Oaks, CA: Sage.
García Coll, C., & Magnusson, K. (1999). Cultural influences
on child development: Are we ready for a paradigm
shift? In A. S. Masten (Ed.), The Minnesota symposia on
child psychology: Vol. 29. Cultural processes in child develop-
ment (pp. 1–24). Mahwah, NJ: Erlbaum.
Gjerde, P. F., & Onishi, M. (2000). In search of theory: The
study of “ethnic groups” in developmental psychology.
Journal of Research on Adolescence, 10, 289–297.
Goodman, S. H., Hoven, C. W., Narrow, W. E., Cohen, P.,
Fielding, B., Alegria, M., Leaf, P. J., Kandel, D., Mccue-
Horowitz, S., Bravo, M., Moore, R., & Dulcan, M. K.
(1998). Measurement of risk for mental disorders and
competence in a psychiatric epidemiological community
survey: The National Institute of Mental Health Meth-
ods for the Epidemiology of Child and Adolescent Men-
tal Disorders (MECA) study. Social Psychiatry and Psychi-
atric Epidemiology, 33, 162–173.
Gottfried, A. E., Fleming, J. S., & Gottfried, A. M. (1998).
Role of cognitively stimulating home environment in
children’s academic intrinsic motivation: A longitudinal
study. Child Development, 69, 1448–1460.
Greenfield, P. M. (1994). Independence and interdepen-
dence as developmental scripts: Implications for theory,
research, and practice. In P. M. Greenfield & R. R. Cock-
ing (Eds.), Cross-cultural roots of minority child development
(pp. 1–37). Hillsdale, NJ: Erlbaum.
Greenfield, P. M. (1995, Winter). Culture, ethnicity, race, and
development: Implications for teaching theory and re-
search. SRCD Newsletter. Chicago, IL: Society for Re-
search in Child Development.
Harkness, S., & Super, C. M. (1995). Culture and parenting.
In M. H. Bornstein (Ed.), Handbook of parenting: Vol. 2. Bi-
ology and ecology of parenting (pp. 211–234). Mahwah, NJ:
Erlbaum.
Harold, G. T., & Conger, R. D. (1997). Marital conflict and
adolescent distress: The role of adolescent awareness.
Child Development, 68, 333–350.
Holmbeck, G. N., Paikoff, R. L., & Brooks-Gunn, J. (1995).
1886 Child Development
Parenting adolescents. In M. H. Bornstein (Ed), Handbook
of parenting: Vol. 1. Children and parenting (pp. 91–118).
Mahwah, NJ: Erlbaum.
Jordan, T. E. (1978). Influences on vocabulary attainment: A
five-year prospective study. Child Development, 49, 1096–
1106.
Kagan, J. (1984). The nature of the child. New York: Basic.
Kelley, M. L., Power, T. G., & Wimbush, D. D. (1992). Corre-
lates of disciplinary practices in low-income Black
mothers. Child Development, 63, 573–582.
Lindahl, K. M., & Malik, N. M. (1999). Marital conflict, fam-
ily processes, and boys’ externalizing behavior in His-
panic American and European American families. Jour-
nal of Clinical Child Psychology, 28, 12–24.
Luster, T., & Dubow, E. (1992). Home environment and ma-
ternal intelligence as predictors of verbal intelligence: A
comparison of preschool and school-age children.
Merrill-Palmer Quarterly, 38, 151–175.
McCall, R. B. (1981). Nature-nurture and the two realms of
development: A proposed integration with respect to
mental development. Child Development, 52, 1–12.
McLoyd, V. C. (1998). Socioeconomic disadvantage and
child development. American Psychologist, 53, 185–203.
McLoyd, V. C. (1999). Cultural influences in a multicultural
society: Conceptual and methodological issues. In A. S.
Masten (Ed.), The Minnesota symposia on child psychology:
Vol. 29. Cultural processes in child development (pp. 123–
135). Mahwah, NJ: Erlbaum.
McLoyd, V. C., Cauce, A. M., Takeuchi, D., & Wilson, L.
(2000). Marital processes and parental socialization in
families of color: A decade review of research. Journal of
Marriage and Family, 62, 1070–1093.
Neiderhiser, J. M., Reiss, D., Hetherington, E. M., & Plo-
min, R. (1999). Relationships between parenting and ado-
lescent adjustment over time: Genetic and environmental
contributions. Developmental Psychology, 35, 680–692.
Parke, R. (1978). Children’s home environments: Social and
cognitive effects. In I. Altman & J. Wohlwill (Eds.), Chil-
dren and the environment (pp. 107–119). New York: Plenum.
Patterson, G. R. (1982). Coercive family processes. Eugene, OR:
Castalia.
Pettit, G. S., Bates, J. E., & Dodge, K. E. (1997). Supportive
parenting, ecological context, and children’s adjustment:
A seven-year longitudinal study. Child Development, 68,
908–923.
Plomin, R., Loehlin, J., & DeFries, J. (1985). Genetic and en-
vironmental components of “environmental” influ-
ences. Developmental Psychology, 21, 391–402.
Rowe, D. C., Jacobson, K. C., & Van den Oord, E. J. C. G.
(1999). Genetic and environmental influences on vocab-
ulary IQ: Parental education level as a moderator. Child
Development, 70, 1151–1162.
Scarr, S. (1992). Developmental theories for the 1990s: Devel-
opment and individual differences. Child Development, 63,
1–19.
Scarr, S., & McCartney, K. (1983). How people make their
own environments: A theory of genotype→environment
effects. Child Development, 54, 424–435.
Scott-Jones, D. (1995). Parent-child interaction and school
achievement. In B. A. Ryan, G. R. Adams, T. P. Gullotta,
R. P. Weisberg, & R. L. Hampton (Eds.), The family–school
connection: Theory, research, and practice (pp. 75–107).
Thousand Oaks, CA: Sage.
Shaw, D. S., Keenan, K., & Vondra, J. I. (1994). Developmen-
tal precursors of externalizing behavior: Ages 1 to 3. De-
velopmental Psychology, 30, 355–364.
Steinberg, L., Darling, N. E., Fletcher, A. C., Brown, B. B., &
Dornbusch, S. M. (1995). Authoritative parenting and
adolescent adjustment: An ecological journey. In P. Moen,
G. H. Elder, & K. Luscher (Eds.), Examining lives in con-
text (pp. 423–466). Washington, DC: American Psycho-
logical Association.
Strauss, M. A. (1991). Discipline and deviance: Physical
punishment of children and violence and other crime in
adulthood. Social Problems, 38, 101–123.
Talbert, J. E., & McLaughlin, M. W. (1999). Assessing the
school environment: Embedded contexts and bottom-up
research strategies. In S. L. Friedman & T. D. Wachs (Eds.),
Measuring environment across the lifespan (pp. 197–228).
Washington, DC: American Psychological Association.
Vuchinich, S., Bank, L., & Patterson, G. R. (1992). Parenting,
peers, and the stability of antisocial behavior in preado-
lescent boys. Developmental Psychology, 28, 510–521.
Wachs, T. D. (1992). The nature of nurture. Newbury Park,
CA: Sage.
Wachs, T. D. (2000). Necessary but not sufficient. Washington,
DC: American Psychological Association.
Wakschlag, L. S., & Hans, S. L. (1999). Relation of maternal
responsiveness during infancy to the development of be-
havior problems in high-risk youth. Developmental Psy-
chology, 35, 569–579.
Whaley, A. L. (2000). Sociocultural differences in the devel-
opmental consequences of the use of physical discipline
during childhood for African Americans. Cultural Diver-
sity and Ethnic Minority Psychology, 6, 5–12.
 ARTICLE

Maternal Disciplinary Practices

in an At-Risk Population
Rebecca R. S. Socolar, MD, MPH; Jane Winsor, BA; Wanda M. Hunter, MPH;
Diane Catellier, MS; Jonathan B. Kotch, MD, MPH

Objectives: To describe maternal discipline of chil-
dren in at-risk families and to determine factors associ-
ated with disciplinary practices.
Design: Cross-sectional survey.
Setting: At-risk families in North Carolina followed up
in a longitudinal study of child maltreatment.
Participants: Maternal caregivers of 7- to 9-year-old chil-
dren with factors at birth that placed them at risk.
Measurements and Results: A total of 186 maternal
caregivers were interviewed. A measure, based on cod-
ing parental responses, was used to assess disciplinary
practices for 5 different misbehaviors. Limit setting was
the most commonly used disciplinary practice for 4 of 5
misbehaviors, with 63% of mothers reporting that this
method generally worked best. Spanking was more likely
used as a secondary response for each misbehavior, when
the primary one had not succeeded. Conversely, teach-
ing or verbal assertion was always less likely as a sec-
ondary response. Teaching or verbal assertion was used
more commonly for lying than for any other misbehav-
ior, limit setting for disobeying, spanking for stealing, and
spanking with an object for being disrespectful. Regres-
sion modeling for the 4 most common disciplinary prac-
tices showed (P,.05) that black race, lack of Aid to Fami-
lies With Dependent Children receipt, more-educated
mothers, and female sex of child were associated with
higher use of teaching or verbal assertion; a biological
father in the home was associated with less use of limit
setting; and black race and report for child maltreat-
ment were associated with more use of mild spanking.
Conclusions: In this sample, limit setting was the most
common disciplinary technique. Disciplinary practices
used varied depending on the type of misbehavior and
other contextual factors, including child, parent, and fam-
ily characteristics.
Arch Pediatr Adolesc Med. 1999;153:927-934

donotwork.Althoughstresshasbeenshown
Editor’s Note: I wonder what the results would be if this study to be related to parenting practices includ-
were repeated focusing on fathers. Would “disrespect” be the stimu- ing discipline,10,11 the disciplinary practices
lus to elicit the most severe response? from population-based samples of at-risk
Catherine D. DeAngelis, MD families have not been well described.
Physical abuse often starts as disci-
pline that gets out of hand,12,13 so it is plau-

From the Departments
of Pediatrics (Dr Socolar),
Maternal and Child Health
(Ms Winsor and Dr Kotch),
Social Medicine (Ms Hunter),
and Biostatistics (Ms Catellier),
University of North Carolina
at Chapel Hill.
D
ISCIPLINARY PRACTICES can
be important to outcomes
such as child self-esteem,1
consciencedevelopment,1,2
aggression,1,3,4 behavior
problems, delinquency 3,6,7 and subsequent
5,6
adult criminal behaviors,3 depression, and
alcoholism.8 There are notable gaps in our
understanding of issues related to discipline.
Thefrequenciesofdisciplinarypracticesaside
from corporal punishment are not well
known, particularly for nonclinical popu-
lations of families and children. Measures
ofdisciplinehavebeeninadequatetodescribe
thebreadthanddepthoftheconstruct.9 Little
is known about what parents choose as sec-
ondary discipline strategies, when the first
sible that physically abusive parents might
have characteristic discipline styles, but this
has not been studied extensively. Existing
studies suggest that physically abusive fami-
lies are more likely to use harsh corporal
punishment,14 coercion,15 and a negative de-
meanor.16 Neglectful parents tend to be un-
responsive and more negative to their chil-
dren.16,17 Since some maltreatment occurs
when initial attempts at discipline cross a
line into psychological or physical abuse, we
This article is also available on our
Web site: www.ama-assn.org/peds.

ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999

927
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
 SUBJECTS AND METHODS
SAMPLE
Subjects included 186 maternal caregivers (hereafter
referred to as mothers) of children aged 7 to 9 years par-
ticipating in a longitudinal study of child maltreatment.
This sample was drawn from a cohort of 788 mother-
infant pairs enrolled in a prospective study39 of child
abuse and neglect in 1985-1987. The initial study,
“Stress, Social Support, and Abuse and Neglect in High
Risk Infants,”39 included mother-infant dyads recruited
from North Carolina hospitals in the immediate postpar-
tum period. Eighty-five percent of the infants in that
study were selected to be “at risk” for adverse social and
medical outcomes. “At-risk” criteria included low birth
weight (,2500 g), young maternal age (,18 years), and
other major medical or social problems diagnosed at
birth (including no prenatal care, significant birth
defects, and congenital conditions or serious neonatal
illness). The remaining 15% of the subjects in the cohort
met none of the “at-risk” criteria. The mother-infant
pairs in both groups were recruited from 37 geographi-
cally diverse counties in North Carolina. The majority of
the mothers in these dyads were black and lived in low-
income households.
By 1991, review of the Central Registry on Child
Abuse and Neglect, a database maintained by the North
Carolina Division of Social Services, revealed that one
third of the 788 infants in the initial study had been
reported to county child protective service agencies as
maltreated. That year, a sample of the cohort was
selected for inclusion in a new longitudinal, multisite
study of the antecedents and outcomes of child abuse
and neglect, the Longitudinal Studies Consortium on
Child Abuse and Neglect (LONGSCAN).40 Funding limi-
tations compelled the LONGSCAN/North Carolina site
to include in its sample only a portion of the subjects
enrolled in the original stress and social support study.
These subjects were selected in order from separate,
computer-generated randomized lists of reported and
nonreported subjects. For each reported child selected, 2
nonreported children matched for sex, race, age, and
poverty level were selected as controls. The final
LONGSCAN sample includes 243 subjects, 83 of whom
had been reported for maltreatment between 6 months
of age and their first LONGSCAN interview.
As part of the LONGSCAN protocol, the North Caro-
lina site collected data annually on all subjects. In-depth,
face-to-face interviews were completed at age 5, 6, and 8
years. This article is based on data collected for age 8 in-
terviews. Owing to attrition (refusals, lack of cooperation,
and inability to locate), 186 of the 243 mothers were in-
terviewed. Statistical comparisons between the original
LONGSCAN sample and the age 8 subsample show no sig-
nificant differences in sociodemographic, risk, or maltreat-
ment variables.
were interested in whether these at-risk parents typically
resorted to more severe strategies for their secondary dis-
cipline techniques.
Factors that affect disciplinary practices and their
outcomes include child, parent, family, community, and
situational characteristics.9 The relationship of some of
ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999
928
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
MEASURES
Disciplinary Practices
Discipline has been defined in various ways.41-44 A distinc-
tion has been made between reactive discipline, done after
the misbehavior, and proactive discipline, done to promote
good behavior.9 To narrow the scope of the study and use a
definition that was readily operationalized, we defined dis-
cipline as the action a parent takes in response to a misbe-
havior. Thus, we focused only on reactive discipline.
Maternal disciplinary practices were assessed with a new
measure designed to prompt respondents for primary and sec-
ondary responses usually used for each of 5 specific child be-
havior problems: disobedience, hitting a younger or smaller
child, disrespect, lying, and stealing. Our approach was de-
signed to orient respondents toward thinking about their dis-
cipline techniques as specific strategies implemented to ad-
dressspecificproblemswithinthecontextofthepast6months.
We used this approach because we were interested in deter-
mining whether mothers use different strategies for different
behavior problems and because we believed it would be easier
for mothers to remember and candidly report their disciplin-
ary practices when the context focused on their child’s mis-
behavior. We were also interested in determining the prac-
tices mothers are likely to use as secondary strategies when
the first are unsuccessful. Finally, we asked what discipline
method works best for the index child most of the time. Re-
sponses were coded by interviewers to 1 of 13 disciplinary cat-
egories that were adapted in part from the work of Trickett
andSusman14 andWebster-StrattonandSpitzer45:(1)nothing—
avoids dealing with the problem; (2) ignoring—a planned
strategy, as opposed to “nothing”; (3) tell someone else, or get
someone else to discipline; (4) empathy; (5) teaching or verbal
assertion (teaching/verbal assertion)—reasoning, giving ex-
planations, giving alternatives, reminding the child of rules
or expectations; (6) limit setting or logical consequences—time-
out, withdrawal of privileges or threat of withdrawal of privi-
leges, removing child from situation, making a child return
a stolen object, making a child apologize, taking away allow-
ance; (7) verbal or symbolic aggression; (8) guilt induction or
embarrassment; (9) mild physical force—spanking with hand;
(10) moderate physical force—hitting or spanking with an ob-
ject, slapping, grabbing, jerking, shaking, throwing an object
at child; (11) severe social isolation; (12) terrorizing; and (13)
severe physical force or restraint. Interrater reliability was as-
sessed by comparison of original codes with a 10% random
sample coded blindly by the consortium coordinator (who
supervises training and monitors data quality).
Child Factors
Data on child age and sex were obtained. Manageability was
measured by asking mothers, “How easy or difficult would
you say it is to manage her/his behavior? Would you say s/he
is (1) easy, (2) average, (3) difficult?” (This variable was di-
chotomized into easy vs average or difficult for analysis.)
these factors with corporal punishment has been well de-
scribed, but for the most part the relationship between
contextual factors and other types of disciplinary prac-
tices has not been well studied.
In her extensive review of the relationship
between gender and disciplinary practices, Steinmetz2
 Mother Factors
The mother’s race, age, and education (highest grade com-
pleted in school, in years) were assessed. Her history of
physical victimization had been assessed previously (1991-
1994) in the longitudinal study (“When you were a child
or teenager . . . were you ever physically hurt by a parent
or someone else—like hit, slapped, beaten, shaken, burned,
or anything like that? (spanking did not count); . . . were
you ever punished or disciplined by someone in such a way
that you were bruised or physically injured?”). Maternal
depression was assessed using the depression subscale of
the Brief Symptom Inventory.46 The partner status (bio-
logical father, other partner, or no partner) and relation-
ship of the mother to the child (biological mother or other
caregiver) were assessed with a family chart that ascer-
tained who lived in the home and relationships to the child.
Attendance at religious or spiritual services was assessed
for the past year (by means of a 6-point Likert scale with
responses ranging from “never” to “more than once a
week”).
Family Factors
Child maltreatment status was determined through a re-
view of the North Carolina Central Registry of Child Abuse
and Neglect to determine if there was any report (substan-
tiated or unsubstantiated) of abuse or neglect of the index
child. Income as percentage of poverty level was deter-
mined by asking about the total income from all sources
after taxes and deductions (measured in $5000 incre-
ments), and the number of people in the household de-
pendent on this income, and calculating the percentage of
the federal poverty level for 1995. Receipt of Aid to Fami-
lies With Dependent Children (AFDC) was determined in
a question about sources of income for the household. So-
cial support was measured with the Social Provisions Scale.47
Neighborhood support was assessed by summing the 4-point
Likert responses to 3 questions (whether the neighbor-
hood was one in which people help each other, watch out
for each other’s children, and can be depended on). The
number of siblings in the household also was determined,
including full, half, and step siblings).
STATISTICAL ANALYSES
Frequencies were determined for each item of the discipline
measure. Reliability of the measure was assessed with k sta-
tistics. Frequencies also were determined for the child, mater-
nal, and family factors, including summative scales for ma-
ternal depression, social support, and neighborhood support.
There were 5 given misbehaviors. Because stealing was
reported infrequently, it was not included in regression mod-
eling, leaving 4 possible misbehaviors: disobedience, fight-
ing, disrespect, and lying. A respondent was included in
regression modeling if she reported that she had actually
dealt with at least 3 of these 4 possible misbehaviors.
found that discipline depended on the sex of the child.
Similarly, Straus and Donnelly4 showed that boys were
hit more often, and more often subjected to verbal
aggression, than girls. 18 Child aggression has been
examined both as a precursor to and an outcome of
corporal punishment, and a positive association is
ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999
929
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
Logistic regression modeling was performed for each dis-
ciplinary practice (that was reported at more than 5%) as
the dependent variable, and the child, mother, and family
factors listed above as the independent factors. This re-
gression accounted for the correlation of observations within
each subject. The resultant odds ratios are the odds of us-
ing the particular disciplinary practice for a given misbe-
havior. Main effects were entered in the model at a = .10,
and were eliminated using stepwise elimination at a = .05.
All possible 2-way interactions with significant main ef-
fects were tested and retained at a = .05.
In addition, we were interested in examining what each
mother chose first as a disciplinary response to a misbe-
havior, ie, the primary response, and what mothers chose
to do when the primary response did not work, ie, the sec-
ondary response. To examine factors associated with dif-
ferent primary responses relative to secondary responses,
we divided the disciplinary techniques into 3 categories:
(1) corporal punishment (codes 9 and higher); (2) teaching/
verbal assertion or limit setting (codes 5 and 6); and (3)
other (codes 1-4, 7, and 8). We then looked at 4 primary
to secondary response groups (since the frequency for the
“other” code was small): corporal punishment (primary)
to corporal punishment (secondary) (CP to CP); corporal
punishment to teaching/verbal assertion or limit setting (CP
to T/LS); teaching/verbal assertion or limit setting to cor-
poral punishment (T/LS to CP); and teaching/verbal asser-
tion or limit setting to teaching/verbal assertion or limit-
setting (T/LS to T/LS). The T/LS to CP response group starts
with nonphysical techniques and moves on to physical
means if the other methods do not change behavior. The
CP to T/LS response group starts with physical methods
and then backs off to less intense methods. When the pri-
mary and the secondary responses are the same (ie, CP to
CP and T/LS to T/LS), this might be thought of as a gauge
of the intensity of use of that type of discipline. Since the
practice is not changing contingent on the child’s behav-
ior, it could potentially be a gauge of noncontingent dis-
ciplinary practices as well.
We examined the frequencies for each misbehavior
for these 4 primary to secondary response groups when
subjects reported them for at least 3 of the 4 given misbe-
haviors. We examined 4 regression models—one each
with the 4 primary to secondary response groups as the
dependent variable, and the same array of independent
variables as above, but found 2 models not to be robust
due to small frequencies. Thus, we report bivariate analy-
sis for all independent variables found to be significant at
a , .05, for the primary to secondary response groups.
Dichotomous variables are reported using odds ratios, and
continuous variables using means for 1 or more misbehav-
iors vs never used.
Data were analyzed using SAS software (SAS Insti-
tute, Cary, NC) on a personal computer. The study was ap-
proved by the University of North Carolina Institutional
Review Board for the School of Public Health, and in-
formed consent was obtained from all participants.
found between aggression and corporal punishment in
both cases.7,19-23
Several studies have shown that older parents are
less likely to use corporal punishment.4,24,25 Multiple stud-
ies4,23,24,26,27 have shown that being spanked as a child is
one of the strongest predictors of whether one would
 ships to disciplinary practices, although each plausibly
Table 1. Characteristics of Families (n = 186)* might affect disciplinary practices.
Several studies indicate that lower socioeconomic
Child status is associated with more corporal punishment, but
Age, y 8.5 (7.8-9.6)
in the research by Straus and Donnelly,4 once the age and
Sex, female 55
Manageability (average or difficult) 59 ethnic group of the parents were taken into account, there
Mother was no longer an association. Several studies have shown
Age, y larger (or more crowded) families to be more likely to
,26 24 use power-assertive methods.1,31,35
26-35 49 The child abuse literature shows a strong associa-
$36 27
tion between stress and lack of social support and physi-
Race
White 34
cal abuse.13,36 Thus, the practice of extreme forms of cor-
Black 66 poral punishment might be related to lack of social support
Education and/or higher stress levels.
Did not complete high school 41 In previous research, Socolar and Stein24,37 found that
Completed only high school 42 the beliefs of parents and physicians about spanking and
Physically victimized as child or teenager 28 other types of discipline depend on whether the disci-
Brief Symptom Inventory score for 2.10 (0-21)
maternal depression
pline is in response to dangerous misbehaviors com-
Relationship to child (biological mother†) 80 pared with annoying ones. Others38 have also found dif-
Partner status ferences depending on the kind of transgression. It is likely
Biological father in home 27 that parents use different disciplinary strategies depend-
Nonbiological father figure in home 17 ing on the type of misbehavior, although this aspect of
No father figure in home 56 the context has largely been ignored to date.
Religious service attendance‡ 2.52 (0-5)
Social support§ 75.50 (56-96)
In previous research about the relationship be-
Family tween disciplinary practices and contextual factors, only
Child reported for abuse or neglect 42 a very limited number of factors have been examined. In
before age 8 y this study, we explored a range of contextual factors likely
Had only abuse report(s) 5 to be associated with disciplinary practices using mod-
Had only neglect report(s) 65 els that allowed us to control for other variables. In ad-
Had both abuse and neglect report(s) 30
Percentage of federal poverty level (1995) 99.7% (11.0%-417.0%)
dition, while there are several studies of disciplinary prac-
Receive AFDC 40 tices in small samples of neglectful and/or abusive
Neighborhood support\ 8.96 (3-12) families,13 research does not exist that focuses on disci-
No. of siblings (full, half, and step) 1.4 (0-7) pline in larger or population-based samples of at-risk fami-
lies. Thus, our goals were to examine the relationship of
*Data are given as mean (range) or percentage unless indicated otherwise. numerous contextual factors to disciplinary practices, to
AFDC indicates Aid to Families With Dependent Children.
†Analysis of the same sample at age 4 of index child revealed that at that
use a new measure of discipline that is more compre-
time 89% of caregivers were the biological mother. hensive and contextually rich, to describe primary and
‡Measured with 6-point Likert scale, ranging from “never” to “more than secondary disciplinary techniques, and to learn about dis-
once a week.” ciplinary practices in a statewide sample of at-risk families.
§Measured with the Social Provisions Scale.
\Measured by summing 4-point Likert scale responses to 3 questions
(whether neighborhood is one in which people help each other, watch out RESULTS
for each other’s children, and can be depended on).
A total of 186 mothers were interviewed. For our sample,
spank as a parent. In addition, parental sex is found to risk factors present at birth included no prenatal care
be associated with disciplinary practice, in that mothers (15%), congenital condition (21%), birth weight less than
were found to spank more often than fathers,4 though 2500 g (32%), and maternal age younger than 18 years
this may be an effect of the greater amount of time that (33%). Multiple demographic and social measures show
mothers spend with children, as compared with fathers. these families to be disadvantaged (Table 1).
There is conflicting evidence about disciplinary styles of Of the 13 discipline codes, only 4 were reported as
black families, but a fair amount of evidence suggests that primary or secondary responses by more than 5% of re-
they tend to be more power assertive, punitive, and ar- spondents: teaching/verbal assertion, limit setting, mild
bitrary.28-31 Elder et al10 and Patterson et al32 showed that physical force, and moderate physical force (Table 2).
stressful experiences increase psychologic distress in par- The agreement between coders was excellent, with an
ents, which in turn increases aversive, coercive, and in- overall k of 0.91 for questions about disciplinary prac-
consistent discipline of children. McLoyd and cowork- tice (95% confidence interval, 0.87-0.96; P,.001).
ers33 found that unemployed mothers had increased Limit setting was the most commonly used disci-
depressive symptoms and that this predicted more fre- plinary practice for 4 of 5 misbehaviors. Sixty-three per-
quent punishment of their adolescent children. McLoyd34 cent of mothers reported that limit setting worked best
linked the disciplinary styles of blacks with the in- with their children most of the time. Teaching/verbal as-
creased stressors they experience—economic and other. sertion was used more commonly for lying than for any
There is not much research regarding parental part- other misbehavior, limit setting for disobeying, mild physi-
ner status, education, and religiosity, and their relation- cal force for stealing, and moderate physical force for be-

ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999

930
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
 Table 2. Use of Disciplinary Practices as Primary Strategy for Misbehaviors

% of Respondents

Teaching Limit Mild Moderate

Misbehavior (No.)* Verbal Assertion Setting Physical Force Physical Force Other
Disobeys (180) 16 60 11 6 7
Fights with younger child (138) 34 43 9 4 10
Is disrespectful (148) 28 32 20 12 8
Lies (149) 46 33 7 7 7
Steals (61) 16 51 25 5 3

*No. varies by misbehavior because responses shown are only for actual misbehavior and discipline used in the past 6 months; no hypothetical responses
are included.

Table 3. Use of Disciplinary Practices as Secondary Strategy for Misbehaviors

% of Respondents

Teaching Limit Mild Moderate

Misbehavior (No.)* Verbal Assertion Setting Physical Force Physical Force Other
Disobeys (163) 4 45 33 12 6
Fights with younger child (119) 4 60 25 7 4
Is disrespectful (127) 5 47 22 14 12
Lies (122) 7 57 18 9 9
Steals (55) 15 31 31 4 19

*No. varies by misbehavior because responses shown are only for actual misbehavior and discipline used in the past 6 months; no hypothetical responses are
included.

ing disrespectful. Mild physical force was more likely as
a secondary response when the primary one had not suc-
ceeded for each misbehavior, whereas teaching/verbal as-
sertion was always less likely as a secondary response
(Table 3). Thirteen percent of respondents reported CP
to CP for at least 1 misbehavior; 33% reported CP to T/LS;
58% reported T/LS to CP, and 73% reported T/LS to T/LS
for at least 1 misbehavior.
Regression models were evaluated using the 4 most
reported disciplinary practices as dependent variables and
child, mother, and family characteristics as the indepen-
dent variables (Table 4). Independent factors that were
predictive of disciplinary technique in multiple models
included race, AFDC receipt, maternal education, and re-
port of maltreatment. Black race was associated with teach-
ing/verbal assertion and mild physical force, and race was
part of an interactive effect with maternal age for limit
setting. Those receiving AFDC were less likely to use
teaching/verbal assertion and more likely to use moder-
ate or severe physical force. Mothers who were more edu-
cated reported significantly more use of teaching/verbal
assertion and less use of moderate or severe physical force.
Mothers who had been reported for maltreatment of the
child were more likely to use mild physical punishment
but less likely to use moderate or severe physical pun-
ishment. Limit setting was used less often in homes with
biological fathers than in homes with no father present,
whereas limit setting was used more with a nonbiologi-
cal father in the home compared with homes with no fa-
ther. Girls were disciplined with teaching/verbal asser-
tion more often than boys. Mothers with a history of
victimization reported that they used mild physical force
less often than nonvictimized mothers.
Bivariate analysis of factors associated with the 4 pri-
mary to secondary response groups showed that the
poorer families use corporal punishment as both a pri-
mary and a secondary response (Table 5). Factors as-
sociated with using corporal punishment first and teach-
ing/verbal assertion or limit setting as a secondary response
were black race, receipt of AFDC, the absence of the bio-
logical father, and less social support. There was an as-
sociation between mothers with more education and more
religious service attendance and use of teaching/verbal
assertion or limit setting as a primary response and cor-
poral punishment as a secondary response. The use of
teaching/verbal assertion or limit setting as primary and
secondary responses was more likely for female chil-
dren. The use of corporal punishment as primary and sec-
ondary responses was more likely for poorer families.
Factors that we studied that were not associated with
any disciplinary type or response group in the analyses per-
formed included child manageability, maternal depres-
sion, relationship of the mother to the child (biological
or other caregiver), and the number of siblings of the child.
COMMENT
In this sample, limit setting was the most common dis-
ciplinary technique reported by mothers. It would be in-
teresting to know to what extent this is generalizable to
other families. Patterson et al48 found that the most fre-
quent parental discipline techniques in their sample of
antisocial boys were ignoring (18%), commanding or re-
questing (15%), giving a time-out (10%), and scolding
or nattering (8%). It is difficult to compare disparate clas-
sification systems for parental discipline, but likely that

ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999

931
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
 Table 4. Logistic Regression Models for 4 Types of Primary Discipline: Factors Associated With These Disciplinary Practices*

Odds Ratio†
Parameter Estimate (95% Confidence Interval)
Teaching verbal assertion
Child’s sex (1 = male, 2 = female) 0.531 1.7 (1.1-2.5)
Race (1 = white, 2 = black) 0.488 1.6 (1.0-2.5)
Maternal education (years of school) 0.122 1.1 (1.0-1.3)
AFDC (0 = no, 1 = receiving) −0.624 0.5 (0.3-0.9)
Religious attendance* ... ...
Neighborhood support* ... ...
Interactions
Religious attendance 3 neighborhood support 0.073 ...
Religious attendance given low neighborhood support NS NS
Religious attendance given high neighborhood support 0.497 1.6 (1.0-2.8)
Limit setting
Biological father in home (0 = no, 1 = yes) −0.623 0.5 (0.3-0.9)
Nonbiological father in home (0 = no, 1 = yes) 0.461 1.6 (1.0-2.6)
Race* ... ...
Maternal age* ... ...
Interactions
Race 3 maternal age 0.057 ...
Maternal age given black race NS NS
Maternal age given white race −0.731 0.5 (0.3-0.9)
Mild physical force
Race (1 = white, 2 = black) 0.755 2.1 (1.1-4.1)
History of maternal victimization (0 = no, 1 = yes) −0.695 0.5 (0.25-0.98)
Report of maltreatment (0 = no, 1 = yes) 0.755 2.1 (1.2-3.7)
Moderate or severe physical force
Maternal education (years of school) −0.385 0.7 (0.6-0.8)
Report of maltreatment (0 = no, 1 = yes) −1.300 0.3 (0.1-0.6)
AFDC (0 = no, 1 = receiving) 0.886 2.4 (1.2-4.9)

*n = 157; a = .05. All factors and 2-way interactions with significant associations are shown. AFDC indicates Aid to Families With Dependent Children; ellipses,
main effects not interpretable due to interactions and NS, not significant.
†Odds ratio is for the odds of using this disciplinary practice for a misbehavior.

commanding or requesting and time-out would fall un-
der limit setting, making it the most common discipline
in the sample of Patterson et al as well. There is very little
literature describing disciplinary practices, other than cor-
poral punishment, for general populations. Such descrip-
tive research would be helpful as a first step toward learn-
ing about the effects of various disciplinary practices in
the general population.
We found that disciplinary practices depend on con-
textual factors including child, parent, family, and situ-
ational characteristics. Increasingly we see that it
does not make sense to ask simply, “What kind of
discipline does a parent use?” But rather, “What kind
of discipline does a given parent use for a given child,
in a specific family for a particular misbehavior?” It is
notable that black mothers in this sample tended to use
teaching or verbal assertion more and to spank more than
white mothers. The finding that they spank more is not
new, but that they use teaching or verbal assertion more
is notable. Black mothers may emphasize their role as
teachers even when they use corporal punishment. It will
be interesting to explore this finding in more detail in
the future.
Aversive stimuli are known to stimulate aggression
in animals and humans.49,50 The most aggressive paren-
tal behavior included in this study, spanking with an ob-
ject, was used for the misbehavior of disrespect. This may
suggest that parents found disrespect to be the most aver-
sive misbehavior. In addition, it raises the question of
whether physical punishment is often a parental reflex-
ive response to aversive stimuli rather than a planned strat-
egy or a response to a child’s need for learning. There is
evidence that parents use mild physical punishment as
a planned strategy,24 but it is not clear if this is the case
for more severe forms of physical punishment.
The factors that we found associated with the 4 pri-
mary to secondary response groups must be considered as
exploratory, since they are bivariate analyses, and do not
control for other independent variables. Using corporal pun-
ishment as a primary response and teaching/verbal asser-
tion or limit setting as a secondary response implies a co-
ercive and potentially illogical contingency pattern for
parental response to child behavior. The association of black
race, AFDC receipt, the absence of the biological father, and
less social support with using this pattern of response for
1 or more misbehaviors needs to be explored further to de-
termine the motivation and efficacy associated with this re-
sponse. Perhaps these mothers are trying to use corporal
punishment to stop the misbehavior and get the child’s at-
tention, before going on to discuss it.
The new measure of discipline used in this study was
helpful in exploring the relationship between the type
of misbehavior and type of discipline, and allowed the
extension of the existing body of research about disci-
pline that focuses on situational context. We also gained
new knowledge about some of the patterns mothers ex-

ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999

932
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
hibit as they follow-up on primary strategies—namely,
that mild physical force is used more often and teaching/ Table 5. Logistic Regression Models: Factors Associated
verbal assertion less often as a backup than as a primary With Primary to Secondary Response Groups*
response to misbehavior. On the other hand, this mea-
Odds Ratio
sure indicated that fewer than 5% of mothers reported (95%
verbal aggression. It is likely that the true incidence of Confidence
yelling is higher than this.18 Possibly, yelling is remem- Response Group Means P Interval)
bered as “talking.” Sometimes there may be a fine dis- CP to CP (15% ever used)
tinction between verbal assertion and verbal aggression. Income as percentage of
Finally, the method of open-ended response with cod- poverty level
ing used in this new measure elicits a richer description Used at least once 61.5
.004 ...
Never used 96.8
of actual parental practices than closed-ended struc-
CP to T/LS (28% ever used)
tured surveys. It would be useful to learn more about Race (black) .005 3.7 (1.4-9.2)
whether an open-ended approach increases the validity AFDC (receiving) .03 2.3 (1.1-5.1)
of measurement of parental discipline compared with a Biological father (present) .02 0.3 (0.1-0.8)
closed-ended approach. Social support (index total:
There are limitations in this study. First, discipline higher = more social support)
Used at least once 72.7 ...
was measured by parental report, rather than actual .01
Never used 76.7
observation. Survey methodology has the advantage that T/LS to CP (53% ever used)
it can be used more readily for large samples, and has Maternal education (years of school)
the potential to be valid since parents know about their Used at least once 11.7 ...
.02
own behavior. However, Patterson et al 48 found that Never used 10.8
parental report was not as accurate as direct observation Religous attendance (0-5 scale:
never to more than once per week)
for most aspects of discipline. In particular, there could
Used at least once 2.8 ...
be a social acceptability reporting bias against the use of Never used 2.0
.02
physical discipline. In fact, we found a paradox that T/LS to T/LS (73% ever used)
could be due to bias in the willingness of mothers to Child’s sex (female) .03 2.4 (1.1-5.4)
report about moderate or severe physical punishment.
We found that families that had been reported to state *n = 120; a = .05. CP indicates corporal punishment; T/LS, teaching or
child protection authorities for abuse or neglect said that verbal assertion or limit setting; ellipses, not applicable; and, AFDC, Aid to
Families With Dependent Children. Means are compared for continuous
they used mild physical punishment more and moderate variables. Odds ratios are used for dichotomous variables. Mothers must
or severe physical punishment less than other families. have responded for at least 3 of 4 misbehaviors for primary and secondary
The latter may be due to fear of reporting this kind of disciplinary practices to be included in the analysis.
punishment when a family has been in trouble with
state authorities in the past. On the other hand, this AFDC teach or use verbal assertion less and use moder-
finding may be true because both neglectful and abusive ate or severe physical force more?
families are included in the maltreatment report cat- Recently, there has been much interest about the ef-
egory, and neglectful families may actually use moderate fects of fathers on family interactions and child out-
or severe physical punishment less. If physical punish- comes.51 In our study, the presence of a biological father
ment is underreported, then the corollary is that other was associated with the group who used teaching/
types of disciplinary practices are likely to be overre- verbal assertion or limit setting as a primary strategy and
ported. Second, it is difficult to know to what popula- used corporal punishment as a secondary strategy. In ad-
tion these findings can be generalized based on the dition, it is interesting to note that our data suggest that
sample selection process. Third, 20% of mothers were it may not be just the presence or absence of a father, but
nonbiological when the child was 8 years old, and 11% whether the father is biological that makes a difference
were nonbiological when the child was aged 4 years. It is in the extent to which limit setting is used to discipline
likely that many important factors have changed over children. By extension, it seems likely that for some child-
time for a significant fraction of these families, including rearing behaviors the role of the father within the house-
the identity of the primary caregiver. Thus, this cross- hold and the biological ties he has to the family may be
sectional view at 8 years does not embody the ways in more important than whether a father is present.
which the at-risk status may have changed over time. As we continue to learn about discipline, it will be
Fourth, we have described disciplinary practices, but not important to learn more about what is effective in the
outcomes that may be related to those practices. It short-term, and the ways in which disciplinary prac-
would be useful to know more about short- and long- tices affect long-term child outcomes such as prosocial
term outcomes of these practices, eg, which strategies behavior, self-esteem, conscience, and aggression. We will
are more effective in stopping the misbehavior, and what also need to understand more about discipline as an in-
are the ultimate effects of these strategies in terms of teractive process through time, between parent and child
how the child is socialized? Finally, we have described within a contextual frame. As we learn about the short-
relationships between contextual factors and disciplin- and long-term effects of disciplinary practices, and the
ary practices, but for these to be useful, it will be impor- processes involved in those practices, we will need to in-
tant to learn in more detail about the nature of these troduce interventions to help parents discipline in the
relationships. For example, why do families receiving ways that bring about the results they desire.

ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999

933
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
Accepted for publication March 4, 1999.
The Longitudinal Studies Consortium on Child Abuse
and Neglect has been supported by grants 90CA1433 and
90CA1467 from the National Center on Child Abuse and
Neglect, Washington, DC. The analyses presented herein were
supported by grant UO1HD30945 from the National Insti-
tute of Child Health and Human Development, Bethesda, Md.
We thank Vincent Dufort for valuable input at several
stages of the analyses, Dot Browne for editing, and Denise
Craig for help in preparation of the manuscript.
Corresponding author: Rebecca R. S. Socolar, MD,
MPH, CB 7225, Community Pediatrics, University of North
Carolina, Chapel Hill, NC 27599-7225 (e-mail: rsocolar
@med.unc.edu).
REFERENCES
1. Becker WC. Consequences of different kinds of parental discipline. In: Review of
Child Development Research. Hoffman ML, Hoffman MW, eds. New York, NY:
Russell Sage Foundation; 1964:169-208.
2. Steinmetz SK. Disciplinary techniques and their relationship to aggressiveness,
dependency, and conscience. In: Contemporary Theories About the Family. New
York, NY: Macmillan Publishing Co; 1979:405-438.
3. McCord J. Some childrearing antecedents of criminal behavior in adult men.
J Pers Soc Psychol. 1979;37:1477-1486.
4. Straus MA, Donnelly DA. Beating the Devil Out of Them: Corporal Punishment
in American Families. New York, NY: Lexington Books; 1994.
5. Forehand RL, McMahon RJ. Helping the Noncompliant Child: A Clinician’s Guide
to Parent Training. New York, NY: Guilford Press; 1981.
6. Patterson GR. Performance models for antisocial boys. Am Psychol. 1986;41:
432-444.
7. Sampson RJ. The family context of juvenile delinquency. In: Crime in the Mak-
ing: Pathways and Turning Points Through Life. Cambridge, Mass: Harvard Uni-
versity Press; 1993:chap 4.
8. Holmes SJ, Robins LN. The role of parental disciplinary practices in the devel-
opment of depression and alcoholism. Psychiatry. 1988;51:24-35.
9. Socolar RRS. A classification scheme for discipline: type, mode of administra-
tion, context. Aggression Violent Behav Rev J. 1997;2:355-364.
10. Elder G, Nguyen T, Caspi A. Linking family hardship to children’s lives. Child Dev.
1985;56:361-375.
11. Patterson G. Stress: a change agent for family process. In: Garmezy N, Rutter
M, eds. Stress, Coping and Development in Children. Baltimore, Md: Johns Hop-
kins University Press. 1988;235-264.
12. Kadushin A, Martin JA. Child Abuse: An Interactional Event. New York, NY: Co-
lumbia University Press; 1981.
13. Panel of Research on Child Abuse and Neglect of the National Research Council. Un-
derstandingChildAbuseandNeglect. Washington,DC:NationalAcademyPress;1993.
14. Trickett PK, Susman EJ. Parental perceptions of child-rearing practices in physi-
cally abusive and nonabusive families. Dev Psychol. 1988;24:270-276.
15. Lorber R, Felton DK, Reid JB. A social learning approach to the reduction of co-
ercive processes in child abusive families: a molecular analysis. Adv Behav Res
Ther. 1984;6:29-45.
16. Burgess RL, Conger RD. Family interaction in abusive, neglectful, and normal
families. Child Dev. 1978;49:1163-1173.
17. Crittenden PM. Social networks, quality of child-rearing, and child develop-
ment. Child Dev. 1985;56:1299-1313.
18. Vissing YM, Straus MA, Gelles RJ, Harrop JW. Verbal aggression by parents and
psychosocial problems of children. Child Abuse Negl. 1991;15:223-238.
19. Larzelere RE. Moderate spanking: model or deterrent of children’s aggression in
the family? J Fam Violence. 1986;1:27-36.
20. McCord J. Parental behavior in the cycle of aggression. Psychiatry. 1988;51:
14-23.
21. Glueck S, Glueck E. Unraveling Juvenile Deliquency. Cambridge, Mass: Harvard
University Press; 1950.
ARCH PEDIATR ADOLESC MED/ VOL 153, SEP 1999
934
Downloaded from www.archpediatrics.com on January 27, 2010
©1999 American Medical Association. All rights reserved.
22. Huesmann LR, Eron LD, Lefkowitz MM, Walder LO. Stability of aggression over
time and generations. Dev Psychol. 1984;20:1120-1134.
23. Sears RR, Maccoby EE, Levin H. Patterns of Child Rearing. Stanford, Calif: Stan-
ford University Press; 1957.
24. Socolar RRS, Stein REK. Spanking infants and toddlers: maternal belief and prac-
tice. Pediatrics. 1995;95:105-111.
25. Becker WC, Peterson DR, Luria Z, Shoemaker DJ, Hellmer LA. Relations of fac-
tors derived from parent-interview ratings to behavior problems of 5 year olds.
Child Dev. 1962;33:509-535.
26. Graziano AM, Namaste KA. Parental use of physical force in child discipline: a
survey of 679 college students. J Interpers Violence. 1990;5:449-463.
27. Muller RT, Hunter JE, Stollak G. The intergenerational transmission of corporal
punishment: a comparison of social learning and temperament models. Child
Abuse Negl. 1995;19:1323-1335.
28. Erlanger HS. Social class and corporal punishment in child rearing: a reassess-
ment. Am Sociol Rev. 1974;39:68-85.
29. Zussman JU. Relationship of demographic factors to parental discipline tech-
niques. Dev Psychol. 1978;14:685-686.
30. Reis J, Barbera-Stein L, Bennett S. Ecological determinants of parenting. Fam
Relations. 1986;35:547-554.
31. Kelley ML, Power TG, Wimbush DD. Determinants of disciplinary practices in
low-income black mothers. Child Dev. 1992;63:573-582.
32. Patterson G, DeBarsyshe B, Ramsey E. A developmental perspective on antiso-
cial behavior. Am Psychol. 1989;44:329-335.
33. McLoyd VC, Jayaratne TE, Ceballo R, Borquez J. Unemployment and work in-
terruption among African American single mothers: effects on parenting and ado-
lescent socioemotional functioning. Child Dev. 1994;65:562-589.
34. McLoyd VC. The impact of economic hardship on black families and children:
psychological distress, parenting, and socioemotional development. Child Dev.
1990;61:311-346.
35. Elder GH, Bowerman CE. Family structure and child-rearing patterns: the effect
of family size and sex composition. Am Sociol Rev. 1963;28:891-905.
36. Egland B, Breitenbucher M, Rosenberg D. A prospective study of the signifi-
cance of life stress in the etiology of child abuse. J Clin Consult Psychol. 1980;
48:195-205.
37. Socolar RRS, Stein REK. Maternal discipline of young children: context, belief,
and practice. J Dev Behav Pediatr. 1996;17:1-8.
38. Catron TF, Masters JC. Mothers’ and children’s conceptualizations of corporal
punishment. Child Dev. 1993;64:1815-1828.
39. Kotch JB, Browne DC, Ringwalt CL, et al. Risk of child abuse or neglect in a co-
hort of low-income children. Child Abuse Negl. 1995;19:1115-1130.
40. Runyan DK, Curtis P, Hunter W, et al. LONGSCAN: a consortium for longitudinal
studies of child maltreatment and the life course of children. Aggression Violent
Behav. 1998;3:275-285.
41. Howard BJ. Discipline in early childhood. Pediatr Clin North Am. 1991;38:1351-
1369.
42. Bettelheim B. Punishment vs. discipline. Atlantic Monthly. November 1985:
51-59.
43. McCormick KF. Attitudes of primary care physicians toward corporal punish-
ment. JAMA. 1992;267:3161-3165.
44. Rosenfeld A, Levine D. Discipline and permissiveness. Pediatr Rev. 1987;8:
209-215.
45. Webster-Stratton C, Spitzer A. Development, reliability, and validity of the daily
telephone discipline interview. Behav Assess. 1991;13:221-239.
46. Derogatis LR, Spencer PM. The Brief Symptom Inventory (BSI): Administration,
Scoring, and Procedures Manual. Baltimore, Md: Johns Hopkins University School
of Medicine; 1982.
47. Cutrona CE. Social support and stress in the transition to parenthood. J Abnorm
Psychol. 1984;93:378-390.
48. Patterson GR, Reid JB, Dishion TJ. Antisocial Boys: A Social Interactional Ap-
proach. Vol 4. Eugene, Ore: Castalia Publishing Co; 1992:67.
49. Ulrich RE, Azrin NH. Reflexive fighting in response to aversive stimulation. J Exp
Anal Behav. 1962;5:511-520.
50. Berkowitz L. External determinants of impulsive aggression. In: Hartup WW, de
Wit J, eds. Origins of Aggression. The Hague, the Netherlands: Mouton; 1978:
147-165.
51. Lamb ME, ed. The Role of the Father in Child Development. 3rd ed. New York,
NY: John Wiley & Sons Inc; 1997.
Family Relations, 56 (January 2007), 80–91. Blackwell Publishing.
Copyright 2007 by the National Council on Family Relations.

The Predictors of Parental Use of Corporal Punishment

Andrew Grogan-Kaylor Melanie D. Otis*

Abstract: Corporal punishment has been the focus of considerable study over the past decade. Some recent research
suggesting that the use of corporal punishment may have significant long-term negative effects on children has
prompted increasing exploration and interest in the issue. We used tobit regression analysis and data from the 2000
National Longitudinal Survey of Youth to examine both the prevalence and the chronicity of spanking in a nation-
ally representative sample of parents. Mother’s characteristics (e.g., age, education) and neighborhood context did
not show a relationship with parental use of corporal punishment. Among parents who used corporal punishment,
being Protestant had a relatively large relationship with its use. Although children’s externalizing behaviors had some
association with parent’s propensity to spank, findings suggest that use of corporal punishment may be better under-
stood as part of a constellation of behaviors relating to a parenting style. Further, findings indicate that it is easier
to predict the incidence of corporal punishment than to predict its frequency of use.

Key Words: corporal punishment, discipline, parent-child relationships, tobit regression.

Debate over appropriate and effective parental disci-
plinary tactics frequently focuses on questions about
corporal punishment. Research suggests that many
parents have spanked their children at some point
(Gershoff, 2002; Grogan-Kaylor, 2004). It is likely
that many parents who use corporal punishment
consider it to be an elicited response to a child’s
problematic behavior. Over the past two decades,
however, an increasingly rigorous body of research
has suggested that corporal punishment contributes
to increases in a broad spectrum of child behavior
problems (Gershoff; Grogan-Kaylor). Indeed, some
important research suggests that measurable out-
comes of parental use of corporal punishment may
persist well into adulthood (MacMillan et al., 1999;
Straus, 2000). This emerging picture of the use of
corporal punishment as a potential contributor to
the continuation of behavior problems suggests the
need for better understanding of the individual
and contextual factors that may relate to corporal
punishment use.
The purpose of the current study was twofold.
Guided by an ecological framework, we first sought
to identify those contextual and individual factors
that predict parents’ propensity to use corporal pun-
ishment. A second goal of the analysis was to exam-
ine more closely the characteristics of those parents
who indicated the use of corporal punishment to
address the question of whether there exist individ-
ual and community differences related to the fre-
quency of use.
Ecological Theory and Corporal
Punishment Utilization
As previously noted, research examining predictors
of corporal punishment has focused primarily on
individual characteristics of parents and children,
and parental expectations about the effectiveness of
the tactic in achieving disciplinary goals. Little atten-
tion, however, has been given to ways that social
context and cultural norms may influence both the
prevalence and the chronicity of corporal punish-
ment. An ecological framework provides both the
tools and the impetus for considering the influence
of environment on parental and child behavior
(Bronfenbrenner, 1979, 1986). Ecological theory

*Andrew Grogan-Kaylor is an Assistant Professor in the University of Michigan School of Social Work, 1080 S University Avenue, Ann Arbor, MI 48109 (agrogan@
umich.edu). Melanie D. Otis is an Associate Professor in the College of Social Work, University of Kentucky, 651 Patterson Office Tower, Lexington, Kentucky
40506-0027 (melanie.otis@uky.edu).
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis
suggests tenets that guided our decision to consider
both individual and community factors in our
efforts to understand the nature of corporal punish-
ment utilization. The model underscores the interre-
latedness of people and their social and physical
worlds. Parenting does not occur in a vacuum.
Instead, parents’ actions are often viewed and judged
by others around them—neighbors, friends, and
community members. Reactions in these settings, as
well as the parents’ own observations of others’
actions, provide information that may or may not
support the use of corporal punishment. Through
these reciprocal relationships, parents may come to
view corporal punishment as the preferred method
of discipline, as inappropriate only under certain cir-
cumstances, or as inappropriate regardless of the cir-
cumstances of the child’s actions.
The current study used an ecological framework
to develop a fuller picture of the nature of corporal
punishment with the inclusion of factors that under-
score the dynamic relationship between people and
their environment. Specifically, parent and child
characteristics, parental expectations, and child behav-
ior, as well as community factors such as region of
the country and neighborhood quality, were consid-
ered collectively.
Literature Review
Parents’ Characteristics
Past research has suggested that mothers are likely
to spank their children more often than fathers
(Nobes, Smith, Upton, & Heverin, 1999; Wolfner
& Gelles, 1993). Frequently, studies have focused
specifically on mothers’ decisions to use corporal
punishment, arguing that mothers are more likely
to be the family disciplinarian by virtue of their
increased level of contact with the child(ren)
(Dietz, 2000; Straus & Donnelly, 1993). How-
ever, with the growing number of working
mothers, the likelihood that mother, rather than
father, may have greater contact with the child has
changed somewhat. Thus, a contemporary explana-
tion for the predominance of female disciplinarians
may be the large number of female-headed house-
holds, which increases the likelihood that mother
may be the disciplinarian, independent of her
attachment to the workforce. In addition to the
increased contact between mothers and children,
81
the growing number of female households has
further significance because of the relationship
between stressful economic conditions and fre-
quency of use of corporal punishment (Giles-Sims,
Straus, & Sugarman, 1995; McLoyd, 1990; Sariola
& Uutela, 1992).
The relationship between socioeconomic status
(SES) and the use of corporal punishment has been
less consistent. Although some studies have linked
low SES to physical abuse of children (Garbarino,
1977; Straus & Gelles, 1986), others have failed to
find a significant relationship between SES and the
use of corporal punishment (Hashima & Amato,
1994; Straus, 2000; Straus & Donnelly, 1993). A
more fruitful approach to understanding the rela-
tionship between individual economic conditions
and the use of corporal punishment may be found
in considering the relationship between economic
factors and stress. To that end, studies have found
significant links between poverty status (i.e., use of
Aid to Families with Dependent Children [AFDC],
unemployment, percent of poverty level) and the
likelihood of using corporal punishment (Giles-Sims
et al., 1995). For example, in their study examining
factors potentially related to the use of corporal pun-
ishment among mothers of 3- to 5-year-olds, Giles-
Sims et al. found that being an AFDC recipient
alone had a significant impact on the frequency of
spanking; however, these authors suggested that
their findings supported a stress theory explanation
of the use of corporal punishment by mothers,
rather than a specific socioeconomic explanation
(Dietz, 2000; Giles-Sims et al.).
Past research has also supported a link between
certain religious affiliations and the likelihood that
a parent will use corporal punishment. In general,
these studies have suggested that Conservative Prot-
estants are more likely to use corporal punishment
as a disciplinary tactic than parents with other reli-
gious affiliations (Ellison, Bartkowski, & Segal,
1996). Ellison (1996) asserted that these findings
may be related to ‘‘religious differences in the
response decision stage’’ where parents assess poten-
tial actions that might be taken to address their
child’s behavior problem. At this stage, it was
argued that individuals from more conservative
fundamentalist religious backgrounds will be more
likely to view corporal punishment not only as a via-
ble option but also potentially as the most appropri-
ate response to the infraction—the ‘‘spare the rod’’
philosophy of childrearing.
82 Family Relations
Volume 56, Number 1
January 2007
Parental Expectations
Parents who view positive parenting as an effective
tool for influencing child behavior may consider
corporal punishment as inappropriate and even anti-
thetical to achieving the desired results. In contrast,
parents who feel their positive efforts will do little to
affect the wanted outcome may be more likely to
use punitive childrearing practices (Luster & Kain,
1987). Further, some parents who support the use
of corporal punishment in specific situations are
aware of the limited effectiveness of the method, yet
feel that some instances justify its use (Graziano,
Hamblen, & Plante, 1996; Holden, Miller, &
Harris, 1999).
Parenting Style
The use of corporal punishment should also be
examined in the larger context of differences in par-
enting styles. To this end, it has been suggested that
spanking may be one part of an overall approach to
parenting that is devoid of positive input from the
parent. This approach is further characterized by
lack of affection and a general climate of aggression,
regardless of whether physical punishment occurs
(Simons, Johnson, & Conger, 1994; Socolar &
Stein, 1995).
Child Characteristics
Research has shown that parents continue to spank
children into their early adolescent years. Although
most parents of preschool-age children (0 – 4 years)
indicate that they have spanked their children, this
rate declines substantially as children age (Frick,
Christian, & Wootton, 1999; Giles-Sims et al.,
1995; Wauchope & Straus, 1990).
Gender has also been shown to relate to the like-
lihood that a child will be spanked (Day, Peterson, &
McCracken, 1998; Lytton & Romney, 1991; Wolfner
& Gelles, 1993). Studies have consistently indicated
that male children are more likely to be spanked.
Additionally, Giles-Sims et al. (1995) found that
regardless of the age of the child, boys are more
likely to be spanked than girls.
Neighborhood Context
Although several studies have examined the relation-
ship between specific parental and child characteris-
tics and the likelihood and frequency of the use of
corporal punishment, contextual factors have been
examined less thoroughly. When they are addressed,
community size and region of the country are
among the most frequently examined contextual fac-
tors. In general, the presumption associated with the
inclusion of these variables has related to cultural
aspects of community that are believed to influence
disciplinary norms. Past research suggests that par-
ents in rural areas and in the South are more likely
to use corporal punishment than those in urban
areas and other regions of the country (Giles-Sims et
al., 1995). In addition to cultural factors, these stud-
ies suggest that mothers in rural areas may have
fewer support resources and less access to/awareness
of information on alternatives to spanking than
mothers in urban areas.
Incidence and Frequency of Spanking:
Where Are the Differences?
As previously noted in some of the extant literature,
whether a parent spanks and how often he or she
spanks may be explained by a variety of factors.
First, use of corporal punishment continues to be
common in the United States, with over 90% of
parents of toddlers indicating that they have spanked
their child at least once (Straus, 2000; Straus &
Stewart, 1999). For instance, Giles-Sims et al.’s
(1995) examination of data from the 1990 National
Longitudinal Survey of Youth (NLSY) on maternal
spanking found that approximately 61% of mothers
of 3- to 5-year-olds had spanked their child in the
past week. In fact, the majority of parents in many
studies indicated that they have spanked their child
at some point in time (Giles-Sims et al.; Straus,
1983); however, this does not necessarily mean that
parents are motivated to spank in response to all dis-
ciplinary concerns.
In addition to differences in whether or not
parents engaged in corporal punishment at all,
there may be important differences in the fre-
quency with which parents employ corporal pun-
ishment. For instance, in Giles-Sims et al.’s (1995)
study, the frequency of spanking ranged from once
(33.1%) to 15 or more times (1.5%), with an
average of three spankings occurring during the
previous week. The wide variation in whether
a parent has spanked their child, and the fre-
quency with which such disciplinary actions are
taken, suggests that a variety of factors may be at
work in predicting the likelihood that a parent will
use corporal punishment.
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis
Summary and Hypotheses
On the basis of our review of the literature, we
examined the following research questions:
1) What are the effects of individual and contextual
factors on the latent propensity to use corporal
punishment?
2) How can such an analysis be extended to under-
stand the effects of individual and contextual fac-
tors on the decision of whether or not to use
corporal punishment?
3) How can such an analysis be extended to under-
stand the effects of individual and contextual
factors on the frequency of use of corporal punish-
ment for those who engage in this practice?
Methods
Analytic Strategy
Unfortunately, standard analytic techniques are not
well suited to the simultaneous examination of the
incidence and frequency of parental use of corporal
punishment. For example, ordinary least squares
regression can be used to treat the number of spank-
ings in the past week as a continuous variable rang-
ing from zero upward. However, a well-known
property of ordinary least squares regression is that
within the linear regression framework, every incre-
ment of a continuous dependent variable is treated
as an equivalent step. Thus, modeling parental use
of corporal punishment in this fashion means that
the transition from no spankings to one spanking is
treated exactly the same as the transition from one
spanking to each additional spanking, thus conflat-
ing the incidence of corporal punishment with the
frequency of corporal punishment.
Additionally, the distribution of corporal punish-
ment in a large sample such as the NLSY raises some
concerns about whether corporal punishment as
a dependent variable may be appropriately modeled
using standard ordinary least squares regression.
Because there are many parents who do not spank
their children on a regular basis, or at all, the value
of corporal punishment in the NLSY data has many
zeroes. Econometricians have long pointed out that
the use of ordinary least squares regression techni-
ques to analyze variables with many zeroes may
83
result in biased and inconsistent parameter estimates
(Greene, 1997).
Indeed, the measurement strategy employed by
the NLSY gives further weight to these concerns.
The NLSY asks parents about the number of times
they have spanked their children in the past week.
Therefore, parents who have spanked their children
in the past month or year, but not in the past week,
are recorded as nonspankers. Arguably, the measures
of corporal punishment in the NLSY may result in a
variable that is left censored by the data collection
process. As discussed in the original formulation by
Tobin (1958) and in an important expository piece
by Roncek (1992), tobit models are admirably suited
to modeling such a left-censored dependent variable.
The analytic strategy, which has come to be
known as the tobit regression, was originally devel-
oped by Tobin (1958) in a study of the yearly
amount of money that households spent on large
purchases. Recognizing that many households do
not spend any money on large purchases in a particu-
lar year, Tobin developed a regression method that
was appropriate for modeling continuous variables
with a large number of zeroes. Over time, Tobin’s
model, which is in some ways statistically a general-
ization of the technique known as probit regression,
has come to be known as Tobin’s probit or, more
simply and more frequently, as the tobit (Greene,
1997).
Tobit analyses have been used in previous empir-
ical work to model contributors to violent or antiso-
cial behavior because of concerns about large
numbers of zero values in the dependent variable, as
well as concerns about censoring mechanisms. For
example, Grogan-Kaylor and Otis (2003) used tobit
regression to examine the contribution of child
abuse and neglect to adult arrest, whereas Jacobs and
O’Brien (1998) used the tobit model to examine the
use of deadly force by police departments. Although
parental use of corporal punishment does not
approach the level of severity of violence examined
in these two research studies, the general principles
of a dependent variable with a distribution that is
skewed toward zero and the possibility of censoring
in the dependent variable render tobit modeling
appropriate.
Furthermore, important work by McDonald and
Moffit (1980) demonstrated a method for decom-
posing, or taking apart, the coefficients of a tobit
model into two separate sets of coefficients. The
first set of coefficients from this decomposition
84 Family Relations
Volume 56, Number 1
January 2007

procedure applies to the question of whether or not Table 1. Descriptive Statistics

study participants have a nonzero value for the Variable
dependent variable. When the dependent variable of
interest is corporal punishment, this set of coeffi- Continuous variables M (SD)
cients thus refers to the incidence of parental use of Number of spankings 0.26 (0.86)
corporal punishment. The second set of coefficients in past week
applies to the question of the value of the dependent Child’s age in months 111.29 (27.75)
variable for those study participants with a nonzero Mother’s years 13.66 (2.36)
value of the dependent variable, which in the case of of education
corporal punishment may be thought of as the fre- Mother’s age in years 38.57 (2.18)
quency of spanking for those parents who actually Cognitive stimulation 99.68 (14.39)
spank. Therefore, the method of decomposing tobit Externalizing behavior 100.97 (13.48)
coefficients proposed by McDonald and Moffit Internalizing behavior 100.19 (13.77)
provides a unique opportunity to model simulta- Neighborhood quality 0.12 (0.61)
neously, yet separately, the predictors of parental use Categorical variables Frequency (%)
of corporal punishment and the frequency with Gender of child
which corporal punishment is employed. Male 426 (53.25)
Before engaging in multivariate analyses, a corre- Female 374 (46.75)
lation matrix of all independent variables was exam- Race of child
ined for evidence of multicollinearity or high Latino 119 (14.88)
correlations between independent variables. No cor- Black 156 (19.5)
relations high enough to warrant concerns about White 525 (65.63)
these issues were detected. Poverty status
Not in poverty 677 (84.63)
In poverty 123 (15.38)
Sample
Religious affiliation
The sample used in this study was derived from Protestant 428 (53.5)
the NLSY. Begun in 1979, the NLSY is an Catholic 231 (28.88)
ongoing nationally representative panel study of Other 65 (8.13)
men and women (Center for Human Resource None 69 (8.63)
Research, 2000) and has been used in hundreds of
publications on children and families. Since 1986,
the NLSY has also collected information every 2
Measures
years on the children born to women in the origi-
nal analysis sample (Center for Human Resource Information on parental usage of corporal punish-
Research) and joined records of mothers and their ment was obtained from the NLSY from a question
children were the data source for this analysis. In that asked parents about the number of times they
some cases, multiple children per mother were had spanked a particular child within the past week.
interviewed as part of the NLSY. In such cases, Respondents were allowed to define spanking as
one child from the family was randomly selected they saw appropriate.
for participation so the analysis would not violate In addition to parental use of corporal punish-
assumptions about the independent and identical ment, the NLSY also provides information on the
distribution of error terms. To make use of recent level of cognitive stimulation that parents provide to
information, data from the 2000 cohort of the children in the home. Originally devised by Bradley,
NLSY were used for this analysis. The final sample Caldwell, and Elardo (1979), the scale measuring
size for this analysis was 800. Descriptive statistics cognitive stimulation includes items on the number
for the sample are provided in Table 1. Because of books within the home, the frequency with which
the decompositions of the tobit parameters that the child is taken on educational outings, and the
are reported in this paper rely on the first tobit amount of time that parents spend talking with the
analysis, all analyses in this paper rely on this final child (Center for Human Resource Research, 2000).
sample of 800. The cognitive stimulation scale is normed to have
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis 85

a mean of 100 and a standard deviation of 15 in the Table 2. Tobit Regression Model
overall NLSY sample (Center for Human Resource Variable Coefficient SE t p
Research).
Information on children’s demographic charac- Child’s gender (male)
teristics such as age, race, and gender were easily Female 0.080 0.326 0.25 .81
obtained from the NLSY. The NLSY also contains Child’s race (White)
a measure of a family’s poverty status. This measure Latino 0.194 0.581 0.33 .74
is computed by the staff at the Center for Human Black 1.231 0.416 2.96 ,.01
Resources Research at the Ohio State University and Child’s age in months 20.042 0.006 27.00 ,.01
is the result of a comparison of a family’s total net Externalizing behavior 0.107 0.016 6.49 ,.01
income for a given year with the poverty standards problems
provided by the U.S. Census for that year. Internalizing behavior 20.048 0.015 23.16 ,.01
The NLSY also contains questions from the problems
Behavior Problems Index (BPI; Zill & Peterson, Poverty status
1986), an instrument designed to assess the level of (not in poverty)
children’s behavior problems. Drawing largely on the In poverty 0.420 0.456 0.92 .36
Child Behavior Checklist (Achenbach & Edelbrock, Mother’s education 20.060 0.074 20.81 .42
1983), the BPI contains a shorter set of items and is Mother’s age 20.094 0.075 21.24 .21
designed to be administered efficiently within the Cognitive stimulation 20.025 0.012 22.03 .04
context of a larger survey. Research indicates that provided by parents
the BPI factors well into two broad subscales of Neighborhood quality 0.352 0.301 1.17 .24
externalizing and internalizing behavior problems Religious affiliation
(Hofferth, Davis-Kean, Davis, & Finkelstein, 1997). (Protestant)
Externalizing behaviors encompass phenomena such Catholic 21.244 0.469 22.65 .01
as whether children are antisocial or have problems Other 20.870 0.603 21.44 .15
with friends, whereas internalizing behaviors include None 21.095 0.658 21.66 .10
anxiety and depression (Center for Human Resource Region (Northeast)
Research, 2001). North Central 20.609 0.551 21.11 .27
South 0.027 0.510 0.05 .96
West 20.289 0.599 20.48 .63
Results Constant 2.394 3.639 0.66 .51
R 2.714 0.204

Tobit Model Coefficients—Parents’ Propensity to Use
Corporal Punishment
All tobit analyses were conducted using STATA sta-
tistical software. The tobit model yielded a number
of interesting results, which are provided in Table 2.
In interpreting coefficients from a tobit regression
model, it is important to remember that these
coefficients are best interpreted as the effects of a set
of independent variables on a latent propensity
(Roncek, 1992)—in this case, the latent propensity
to use corporal punishment.
Only some of children’s demographic character-
istics were associated with corporal punishment.
Specifically, a child’s gender was not associated with
whether or not her or his parents used corporal pun-
ishment. Also, when compared with parents of
White children, parents of Black children were more
likely to use corporal punishment; however, parents
Note: Reference categories are in parentheses.
of Latino children were no more likely to use corpo-
ral punishment than parents of White children.
Children’s age was negatively associated with the use
of corporal punishment because parents of older
children were less likely to use corporal punishment.
Many family characteristics were not associated
with parental use of corporal punishment. For
example, a mother’s education and age were not
associated with the use of corporal punishment.
However, higher levels of cognitive stimulation pro-
vided in the home was associated with lower levels
of corporal punishment. Religious affiliation was
associated with parental use of corporal punishment
and generated the largest effect sizes of any variables
in the model. Protestant parents were most likely to
use corporal punishment; in comparison, Catholic
86 Family Relations
Volume 56, Number 1
January 2007

parents were less likely to use corporal punishment. Table 3. Decomposition of Tobit Coefficients: Probability
Parents who reported that they had no religious That Child Was Spanked
affiliation did not differ from Protestant parents in Variable dF/dx (%) SE z p
proclivity to use corporal punishment.
There were interesting relationships between Child’s gender (male)
children’s behavior and the use of corporal punish- Female 0.49 0.020 0.25 .81
ment. Children who displayed greater amounts of Child’s race (White)
externalizing behavior problems were more likely to Latino 1.22 0.035 0.34 .73
be the recipients of corporal punishment, whereas Black 8.95 0.025 3.53 ,.01
children who displayed greater levels of internalizing Child’s age in months 20.25 0.000 27.00 ,.01
behavior problems were less likely to be the recipi- Internalizing behavior 20.29 0.001 23.16 ,.01
ents of corporal punishment. In interpreting the problems
effect sizes of these coefficients, it is worth recalling Externalizing behavior 0.65 0.001 6.49 ,.01
that the measurement of children’s behavior was problems
drawn from the BPI (Zill & Peterson, 1986) in Poverty status
which data were normed so that the standard devia- (not in poverty)
tion of both the internalizing and the externalizing In poverty 2.74 0.028 0.99 .32
behavior problems scales was set at 15. Mother’s education 20.36 0.004 20.81 .42
Some larger community-level variables were also Mother’s age 20.57 0.005 21.24 .21
included in the model. Neither the characteristics of Cognitive stimulation 20.15 0.001 22.03 .04
the neighborhood nor the geographic region in Religion (Protestant)
which the family lived showed a relationship with Catholic 26.68 0.029 22.34 .02
parental use of corporal punishment. Other 24.41 0.037 21.20 .23
None 25.31 0.040 21.33 .19
Region (Northeast)
First Decomposition of Tobit Parameters—Effects North Central 23.47 0.034 21.04 .30
of Independent Variables on Use Versus Nonuse South 0.16 0.031 0.05 .96
of Corporal Punishment West 21.69 0.036 20.46 .64
In the second step of the analysis, the coefficients of Neighborhood quality 2.14 0.018 1.17 .24
the tobit model were decomposed into two more Constant 14.57 0.221 0.66 .51
informative sets of parameters using the procedure
Note: Marginal effects: probability uncensored. Incidence: probability that child was
outlined by McDonald and Moffit (1980). Statisti- spanked at least once. Reference categories are in parentheses. dF/dx ¼ change in out-
cally, this ‘‘decomposition’’ may be seen as a process come associated with change in x.
of disentangling—or taking apart—these estimates
into two sets of estimates that are more intuitively a parent would use corporal punishment. Higher
meaningful. The first set of secondary parameters levels of children’s externalizing behavior problems
indicated the extent to which each independent vari- were associated with an increased probability of
able had a relationship with the probability that using corporal punishment, whereas higher levels of
a particular parent would or would not use corporal children’s internalizing behavior problems were
punishment (see Table 3). Parents of Black children associated with a lower probability of corporal pun-
were almost 9% more likely to use corporal punish- ishment. Higher levels of cognitive stimulation were
ment than were White parents. The probability of associated with lower probabilities that corporal
parents using corporal punishment decreased by punishment would be used. However, neither
a quarter of a percent for every month the child aged mothers’ education nor age was associated with cor-
resulting in a 4% reduction in the probability that poral punishment. The internalizing behavior prob-
corporal punishment would be used for every addi- lems score, externalizing behavior problems score,
tional year in children’s age. However, poverty status and cognitive stimulation score were all normed to
did not have an association with whether or not par- have standard deviations of 15. Multiplying the
ents used corporal punishment. parameter estimates for each of these independent
Both internalizing and externalizing behavior variables by 15 suggested that a standard deviation
problems were associated with the probability that change in these quantities was associated with
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis 87

a 4.4% decrease, 9.7% increase, and 2.3% decrease Table 4. Decomposition of Tobit Coefficients: Effect on
in the use of corporal punishment, respectively. Number of Uses of Corporal Punishment for Parents Who
As expected, religion was related to the use of Employed Corporal Punishment
corporal punishment. Catholic parents were almost Variable dF/dx SE z p
7% less likely to use corporal punishment than Prot-
estant parents. But neither geographic region nor Child’s gender (male)
neighborhood quality was associated with corporal Female 0.013 0.054 0.25 .81
punishment usage when accounting for the presence Child’s race (White)
of other variables in the model. Latino 0.032 0.096 0.34 .74
Black 0.220 0.069 3.21 ,.01
Second Decomposition of Tobit Parameters—Effects Child’s age in months 20.007 0.001 27.00 ,.01
of Independent Variables on Frequency of Use Internalizing behavior 20.008 0.003 23.16 ,.01
of Corporal Punishment problems
A second set of parameters was also obtained from Externalizing behavior 0.018 0.003 6.49 ,.01
the decomposition of the tobit coefficients. This sec- problems
ond set of parameters indicated the relationship of Poverty status
each of the independent variables on the number of (not in poverty)
spankings used by parents in the past week, condi- In poverty 0.071 0.075 0.95 .34
tional on those parents having used one or more cor- Mother’s education 20.010 0.012 20.81 .42
poral punishments. Table 4 provides data on this Mother’s age 20.015 0.012 21.24 .21
second set of parameters. Again, children’s gender did Cognitive stimulation 20.004 0.002 22.03 .04
not have an association with the number of times that Religion (Protestant)
corporal punishment was employed, nor did a child’s Catholic 20.195 0.077 22.52 .01
Latino ethnicity. On the other hand, Black children Other 20.133 0.099 21.34 .18
received 0.22 more spankings than White children in None 20.165 0.108 21.52 .13
families where corporal punishment was employed. Region (Northeast)
As expected, every month increase in a child’s age was North Central 20.098 0.091 21.08 .28
associated with a decrease in the number of spankings South 0.004 0.084 0.05 .96
employed, resulting in a 0.12 decrease in the number West 20.047 0.099 20.47 .64
of spankings used for every year that the child aged. Neighborhood quality 0.058 0.050 1.17 .24
In summary, children’s behavior problems were Constant 0.394 0.599 0.66 .51
related to the number of uses of corporal punishment
Note: Marginal effects: conditional on being uncensored. Frequency: effect of inde-
in families that employed corporal punishment. As pendent variables on number of spankings, given that child is spanked. Reference cate-
with the previous analyses, higher levels of children’s gories are in parentheses. dF/dx ¼ change in outcome associated with change in x.
externalizing behaviors were associated with a higher
likelihood of parental use of corporal punishment.
Religion had a relatively large relationship with the
Conversely, children who demonstrated greater levels
frequency of corporal punishment among parents
of internalizing behavior problems were less likely to
employing physical punishment; however, compared
experience corporal punishment. The relationship of
with Protestant parents, Catholic parents spanked
externalizing behavior problems with the use of cor-
much less frequently. As expected, geographic region
poral punishment was greater than the relationship of
and neighborhood quality did not have an association
internalizing behavior problems with the use of cor-
with the number of times parents who used corporal
poral punishment.
punishment actually spanked their children.
As expected, many characteristics of mothers
were not associated with whether or not children
experienced corporal punishment. Mother’s age and
Discussion
mother’s level of education were not associated with
higher levels of use of corporal punishment among
parents who used corporal punishment. Higher levels Concern over the parental use of corporal punish-
of cognitive stimulation were associated with lower ment has been elevated by a growing body of litera-
levels of the use of corporal punishment. ture linking the disciplinary tactic to a number of
88 Family Relations
Volume 56, Number 1
January 2007
negative child welfare outcomes (Gershoff, 2002;
Grogan-Kaylor, 2004; Straus, 2000). In this study,
we sought to expand our understanding of corporal
punishment by using an ecological framework to
examine variation in both the rates and the chronicity
of corporal punishment among a random sample of
parents. Our analyses provide only partial support for
our model, with race and ethnicity, religion, child’s
behavior, and parental use of cognitive stimulation
being the most consistent predictors of variation in
corporal punishment frequency and chronicity.
An important limitation of cross-sectional
research of this nature is an inability to ascertain
causal direction. There are also some other limita-
tions to this research. First, the NLSY simply asks
survey respondents about the number of times that
they have spanked their child within the past week.
Thus, it is left up to respondents to define the mean-
ing of spanking for them. Additionally, although the
NLSY data contain important information about
the incidence and frequency of corporal punish-
ment that forms the basis of this analysis, important
information on the severity with which corporal
punishment is applied is not present in the NLSY
data.
Bearing these limitations in mind, our analysis is
consistent with past research showing that, in com-
parison with White families, mothers in African
American families were more likely to employ corpo-
ral punishment (Giles-Sims et al., 1995; McLoyd &
Smith, 2002). Interestingly, the inclusion of a var-
iable for poverty into the model did not eliminate
the statistical significance of race, indicating that fac-
tors other than economic status may play a role in
the degree to which parents make use of corporal
punishment as a disciplinary strategy. Horn, Cheng,
and Joseph’s (2004) examination of disciplinary
practices among lower SES and middle-to-upper
SES African American parents reported similar find-
ings, with few SES-related differences noted in their
sample of 175 African American parents.
Similarly, in this analysis, the inclusion of mea-
sures of neighborhood quality does not reduce the
relationship of family race or ethnicity with parental
use of corporal punishment. Indeed, the measure of
neighborhood quality is not predictive of parental
use of corporal punishment, which stands in con-
trast with other research exploring this relationship
(Giles-Sims et al., 1995). To some extent, this lack
of statistical significance may be a function of the
fact that the neighborhood scale in the NLSY has
only five items that correlate highly with each other
and appear to represent a single dimension of neigh-
borhood quality.
Other research, which has relied on data with
measures of multiple dimensions of neighborhood
quality has suggested that adequately accounting
for the characteristics of a neighborhood may do
a great deal to explain racial and ethnic disparities in
child outcomes (Woolley & Grogan-Kaylor, 2006).
Woolley and Grogan-Kaylor explored the effect of
family and neighborhood characteristics on three
school-related outcomes and found that in some
cases the inclusion of neighborhood characteristics
in a statistical model helped to fully explain racial
and ethnic disparities in academic outcomes. This
finding calls to mind the assertion in seminal articles
by Massey (1990, 1994) that African American fam-
ilies are often likely to live in very different neigh-
borhoods from White families. Unfortunately, the
data set used by Woolley and Grogan-Kaylor did
not contain measures of parental discipline, render-
ing direct tests of these ideas impossible. More
research on parental discipline that makes use of
data with rich neighborhood indicators is likely to
be fruitful.
Our findings also suggest that mothers’ decisions
on whether or not to employ corporal punishment
are associated with their child’s behavior. Notably,
the present research suggests that even when other
factors are included in a statistical model, children’s
actual behavior is only one of several factors associ-
ated with parental use of corporal punishment.
Indeed, when the parameter estimates associated
with changes in children’s behavior problems were
multiplied by 15 to obtain an estimate of the
changes in parental corporal punishment related to
a standard deviation change in behavior problems,
several other factors in the model had comparable
effect sizes. These findings suggest that cultural and
religious factors are likely to play a significant role in
affecting whether or not parents employ corporal
punishment. Unfortunately, the NLSY data do not
contain finely grained measures of respondents’ reli-
gious beliefs, rendering further exploration of the
religion measures in this analysis impossible. Further
research of this nature is likely to benefit from the
use of data sets that contain more sophisticated mea-
sures of religion than those present in the NSLY.
Similar to other research suggesting that parental
use of corporal punishment may be part of a pattern
of parenting characterized by less use of positive
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis
parenting tactics (Grogan-Kaylor, 2004), the results
of this research provide evidence that parental use of
corporal punishment is associated with a lower inci-
dence of parental provision of a cognitively stimulat-
ing environment for children. Such a finding
suggests that corporal punishment may be part of
a parenting style with lesser amounts of positive par-
enting practices. Given the effect size of parental use
of cognitive stimulation on parental use of corporal
punishment, and the fact that increased amounts of
cognitive stimulation are associated with decreased
amounts of children’s behavior problems (Grogan-
Kaylor, 2005a, 2005b), it may be beneficial to focus
on interventions that teach parents to increase the
amount of intellectual stimulation in the home.
Interestingly, community-level variables included
in the model had no measurable relationship with
parental use of corporal punishment. Family neigh-
borhood and the region of the country in which the
family lived were not associated with parental use of
corporal punishment. Important research has sug-
gested connections between neighborhood condi-
tions and other forms of family violence (Molnar,
Miller, Azrael, & Buka, 2004). There are also peri-
odic suggestions in the literature on parental disci-
pline that corporal punishment may be adaptive in
some kinds of neighborhoods (Eamon, 2001),
although these ideas have been further examined
and found wanting (Grogan-Kaylor, 2005b). This
research finds no evidence that neighborhood condi-
tions are likely to play a role in the decision to use
corporal punishment.
Finally, the decomposition of the tobit coeffi-
cients suggests that it is easier to predict the inci-
dence of parental use of corporal punishment than
the frequency with which parents employ corporal
punishment. The effect sizes of the parameters asso-
ciated with incidence were much larger than those
associated with frequency. For example, the coeffi-
cient for the effect of externalizing behavior prob-
lems on the use of corporal punishment (Table 3)
was 0.65%. Given that the standard deviation for
this measure is approximately 15, an increase of
a standard deviation in children’s externalizing
behavior is associated with approximately a 9.75%
increase in the likelihood of parents using corporal
punishment. Conversely, the coefficient for the
effect of externalizing behavior problems on parental
use of corporal punishment for those parents who
do employ corporal punishment was 0.018. Apply-
ing the same multiplier of 15 indicates that for
89
parents who spank, a standard deviation increase in
children’s externalizing behavior problems is associ-
ated with an additional 0.27 uses of corporal pun-
ishment. Thus, interventions targeted at changing
parents’ use of corporal punishment may have more
effect on parents’ decision on whether or not to use
corporal punishment, rather than on the frequency
with which parents who already make use of corpo-
ral punishment apply this tactic.
Implications for Practice and Policy
Collectively, these findings suggest the need to con-
sider a complexity of factors when attempting to
address the use of corporal punishment by parents.
First, family and children’s advocates wishing to
reduce parental use of corporal punishment would
be well advised to direct attention to the role of
broad cultural factors that have an effect on parents’
valuation of the appropriateness and effect of corpo-
ral punishment. Many parents make a conscious
decision to use corporal punishment on the basis of
beliefs about its appropriateness and effectiveness as
a disciplinary tactic. If practitioners presume the
decision is guided by a desire to be a good parent,
then awareness of more effective alternatives may
provide the necessary information to encourage
alternative disciplinary choices. Efforts to teach par-
ents how to be more effective disciplinarians will not
only acknowledge their desire to be good parents
but also include attention to other key aspects of the
parent-child relationship (American Academy of
Pediatrics, 1998).
Acknowledging the importance of cultural con-
text and the dynamic interaction between individu-
als, families, and communities is also critical to
efforts to affect the utilization of corporal punish-
ment. Individual parenting classes that teach parents
to use nonphysical disciplinary tactics may have little
impact in an environment where corporal punish-
ment is the norm and ‘‘spare the rod, spoil the
child’’ is a guiding principle of parenting. Family
and child advocates must attend to these cultural
differences while also being cognizant of the need
for cultural sensitivity to provide effective program-
ming and reduce barriers to client engagement
(Kalyanpur & Harry, 1997; Lieberman, 1998).
Future research should continue to examine the
links between individual factors and social and cul-
tural factors that are associated with the utilization
90 Family Relations
Volume 56, Number 1
January 2007
of corporal punishment. Although the current study
only partially supports the relevance of environmen-
tal factors, the findings do suggest the utility of an
ecological framework for guiding future efforts to
clarify these relationships. Thus, future research
should consider the complex array of cultural norms
and values that influence parenting beliefs and
behaviors to better understand the interrelationship
of these factors.
References
Achenbach, T., & Edelbrock, C. (1983). Manual for the child behavior
checklist and revised child behavior profile. Burlington: University of Ver-
mont, Department of Psychology.
American Academy of Pediatrics. (1998). Guidance to effective discipline.
Pediatrics, 101, 723–728.
Bradley, R. H., Caldwell, B. M., & Elardo, R. (1979). Home environment
and cognition development in the first two years: A cross-lagged panel
analysis. Developmental Psychology, 15, 246–250.
Bronfenbrenner, U. (1979). The ecology of human development. Cambridge,
MA: Harvard University Press.
Bronfenbrenner, U. (1986). Ecology of the family as a context for human
development: Research perspectives. Development Psychology, 22, 723–742.
Center for Human Resource Research. (2000). NLSY79: 1998 child and
young adult data user’s guide (The Ohio State University). Columbus:
The Ohio State University, Center for Human Resource Research.
Center for Human Resource Research. (2001). NLSY79 user’s guide: A guide
to the 1979–2000 National Longitudinal Survey of Youth data. Colum-
bus: The Ohio State University, Center for Human Resource Research.
Day, R. D., Peterson, G. W., & McCracken, C. (1998). Predicting spank-
ing of younger and older children by mothers and fathers. Marriage and
Family, 60(1), 79–94.
Dietz, T. L. (2000). Disciplining children: Characteristics associated with the
use of corporal punishment. Child Abuse & Neglect, 24, 1529–1542.
Eamon, M. K. (2001). Poverty, parenting, peer and neighborhood influen-
ces on young adolescent antisocial behavior. Journal of Social Service
Research, 28(1), 1–23.
Ellison, C. G., Bartkowski, J. P., & Segal, M. L. (1996). Conservative Prot-
estantism and the parental use of corporal punishment. Social Forces,
74, 1003–1028.
Frick, P. J., Christian, R. E., & Wootton, J. M. (1999). Age trends in the
association between parenting practices and conduct problems. Behav-
ior Modification, 23(1), 106–128.
Garbarino, J. (1977). The human ecology of child maltreatment: A concep-
tual model for research. Journal of Marriage and the Family, 39, 721–732.
Gershoff, E. T. (2002). Corporal punishment by parents and associated
child behaviors and experiences: A meta-analytic and theoretical review.
Psychological Bulletin, 128, 539–579.
Giles-Sims, J., Straus, M. A., & Sugarman, D. B. (1995). Child, maternal,
and family characteristics associated with spanking. Family Relations,
44(2), 70–176.
Graziano, A. M., Hamblen, J., & Plante, W. A. (1996). Subabusive vio-
lence in child-rearing in middle class American families. Pediatrics, 98,
845–848.
Greene, W. H. (1997). Econometric analysis (3rd ed.). Upper Saddle River,
NJ: Prentice Hall.
Grogan-Kaylor, A. (2004). The effect of corporal punishment on antisocial
behavior in children. Social Work Research, 28(3), 153–164.
Grogan-Kaylor, A. (2005a). Corporal punishment and the growth trajectory
of children’s antisocial behavior. Child Maltreatment, 10, 283–292.
Grogan-Kaylor, A. (2005b). Relationship of corporal punishment and anti-
social behavior by neighborhood. Archives of Pediatrics & Adolescent
Medicine, 159, 938–942.
Grogan-Kaylor, A., & Otis, M. (2003). The effect of child maltreatment
on adult criminal behavior: A tobit regression analysis. Child Maltreat-
ment, 8(2), 129–137.
Hashima, P. Y., & Amato, P. R. (1994). Poverty, social support, and paren-
tal behavior. Child Development, 65, 394–403.
Hofferth, S., Davis-Kean, P. E., Davis, J., & Finkelstein, J. (1997). The
child development supplement to the panel study of income dynamics:
1997 user guide. Retrieved February 15, 2005, from http://psidonline.
isr.umich.edu/CDS/usergd.html
Holden, G. W., Miller, P. C., & Harris, S. D. (1999). The instrumental
side of capital punishment: Parents’ reported practices and outcome
expectancies. Journal of Marriage and Family, 61, 908–919.
Horn, I. B., Cheng, T. L., & Joseph, J. (2004). Discipline in the African
American community: The impact of socioeconomic status on beliefs
and practices. Pediatrics, 115, 1236–1241.
Jacobs, D., & O’Brien, R. M. (1998). The determinants of deadly force: A
structural analysis of police violence. American Journal of Sociology,
103, 837.
Kalyanpur, M., & Harry, B. (1997). A posture of reciprocity: A practical
approach to collaboration between professionals and parents of culturally
diverse backgrounds. Journal of Child and Family Studies, 6, 487–509.
Lieberman, A. F. (1998). Culturally sensitive intervention with children and
families. Newsletter of the Infant Health Promotion Project, 22, 1–5.
Luster, T. & Kain, E. L. (1987). The relation between family context and
perceptions of parental efficacy. Early Childhood Development and Care,
29(3), 301–311.
Lytton, H., & Romney, D. M. (1991). Parents’ differential socialization of
boys and girls: A meta-analysis. Psychological Bulletin, 109, 267–296.
MacMillan, H. L., Boyle, M. H., Wong, M. Y., Duku, E. K., Fleming,
J. E., & Walsh, C. A. (1999). Slapping and spanking in childhood and
its association with lifetime prevalence of psychiatric disorders in a gen-
eral population sample. Canadian Medical Association Journal, 161,
805–810.
Massey, D. S. (1990). American apartheid: Segregation and the making of
the underclass. The American Journal of Sociology, 96(2), 329–357.
Massey, D. S. (1994). America’s apartheid and the urban underclass. The
Social Service Review, 68(4), 471–478.
McDonald, J. F., & Moffit, R. A. (1980). The uses of tobit analysis. Review
of Economics and Statistics, 62, 318–321.
McLoyd, V. C. (1990). The impact of economic hardship on Black families
and children: Psychological distress, parenting, and socioemotional
development. Child Development 61, 311–346.
McLoyd, V. C., & Smith, J. (2002). Physical discipline and behavior prob-
lems in African American, European American, and Hispanic children:
Emotional support as a moderator. Journal of Marriage and Family,
64(1), 40–53.
Molnar, B. E., Miller, M. J., Azrael, D., & Buka, S. L. (2004). Neighbor-
hood predictors of concealed firearm carrying among children and ado-
lescents: Results from the project on human development in Chicago
neighborhoods. Archives of Pediatric and Adolescent Medicine, 158,
657–664.
Nobes, G., Smith, M., Upton, P., & Heverin, A. (1999). Physical punish-
ment by mothers and fathers in British homes. Journal of Interpersonal
Violence, 14, 887–902.
Roncek, D. W. (1992). Learning more from tobit coefficients: Extending
a comparative analysis of political protest. American Sociological Review,
57, 503.
Sariola, H., & Uutela, A. (1992). The prevalence and context of family vio-
lence against children in Finland. Child Abuse & Neglect, 16, 823–832.
Simons R. L., Johnson, C., & Conger, R. D. (1994). Harsh corporal punish-
ment versus quality of parental involvement as an explanation of adoles-
cent maladjustment. Journal of Marriage and the Family, 56, 591–607.
Socolar, R. R., & Stein, R. E. (1995). Spanking infants and toddlers:
Maternal belief and practice. Pediatrics, 95(1), 105–111.
Straus, M. A. (1983). Ordinary violence, child abuse, and wife beating:
What do they have in common? In D. Finkelhor, R. Gelles, G. Hotal-
ing, & M. Straus (Eds.), The dark side of families: Current family violence
research (pp. 213–234). Beverly Hills, CA: Sage.
Straus, M. A. (2000). Beating the devil out of them: Corporal punishment in
American families and its effects on children. New Brunswick, NJ: Trans-
action Publications.
Straus, M. A., & Donnelly, A. D. (1993). Corporal punishment of adoles-
cents by American parents. Youth and Society, 24, 419–420.
 The Predictors of Parental Use of Corporal Punishment
Grogan-Kaylor and Otis
Straus, M. A., & Gelles, R. J. (1986). Societal change and change in family
violence from 1975 to 1985 as revealed in two national surveys. Journal
of Marriage and the Family, 48, 465–479.
Straus, M. A., & Stewart, J. H. (1999). Corporal punishment by American
parents: National data on prevalence, chronicity, severity, and duration,
in relation to child, and family characteristics. Clinical Child and Family
Psychology Review, 2(2), 55–70.
Tobin, J. (1958). Estimation of relationships for limited dependent varia-
bles. Econometrica, 26(1), 24–36.
Wauchope, B., & Straus, M. A. (1990). Physical punishment and
physical abuse of American children: Incidence rates by age, gender,
91
and occupational class. In M. Straus & R. Gelles (Eds.), Physical vio-
lence in American families: Risk factors and adaptations to violence
in 8,145 families (pp. 133–148). New Brunswick, NJ: Transaction
Publications.
Wolfner, G. D., & Gelles, R. J. (1993). A profile of violence toward chil-
dren: A national study. Child Abuse & Neglect, 17, 197–212.
Woolley, M. E., & Grogan-Kaylor, A. (2006). Protective family factors in
the context of neighborhood: Promoting positive school outcomes.
Family Relations, 55(1), 93–104.
Zill, N., & Peterson, J. (1986). Behavior problems index. Washington, DC:
Child Trends.