FOR UNICEW -Research Essay
Olivia
wolfe_crooks_lee_2.pdf
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Clinical Child and Family Psychology Review, Vol. 6, No. 3, September 2003 ( C© 2003)
The Effects of Children’s Exposure to Domestic Violence: A Meta-Analysis and Critique
David A. Wolfe,1,2,4 Claire V. Crooks,1 Vivien Lee,1
Alexandra McIntyre-Smith,1 and Peter G. Jaffe3
A wide range of children’s developmental outcomes are compromised by exposure to domestic violence, including social, emotional, behavioral, cognitive, and general health functioning. However, there are relatively few empirical studies with adequate control of confounding variables and a sound theoretical basis. We identified 41 studies that provided relevant and adequate data for inclusion in a meta-analysis. Forty of these studies indicated that children’s exposure to domestic violence was related to emotional and behavioral problems, translating to a small overall effect (Zr = .28). Age, sex, and type of outcome were not significant moderators, most likely due to considerable heterogeneity within each of these groups. Co-occurrence of child abuse increased the level of emotional and behavioral problems above and beyond exposure alone, based on 4 available studies. Future research needs are identified, including the need for large-scale longitudinal data and theoretically guided approaches that take into account relevant contextual factors.
KEY WORDS: domestic violence; child witnesses; marital aggression; meta-analysis; child behavior prob- lems; family violence; child abuse.
Describing the effects of exposure to domestic vi- olence on children and adolescents has been the sub- ject of intense research efforts since early studies on this topic emerged in the mid-1980s. Several scholarly reviews have been conducted of this literature, result- ing in a general consensus that exposure to domes- tic violence has a significant and measurable negative effect on children’s functioning, relative to children from nonviolent families (Edleson, 1999; Fantuzzo & Lindquist, 1989; Fantuzzo & Mohr, 1999; Margolin & Gordis, 2000; Wolak & Finkelhor, 1998). These
1Centre for Research on Violence Against Women and Children, The University of Western Ontario, London, Ontario, Canada.
2Centre for Addiction and Mental Health, The University of Toronto, Toronto, Canada.
3Centre for Children and Families in the Justice System of the London Family Court Clinic, London, Ontario, Canada.
4Address all correspondence to David A. Wolfe, Centre for Research on Violence Against Women and Children, The Uni- versity of Western Ontario, 1137 Western Road, Room 1118, Fac- ulty of Education Building, London, Ontario, Canada N6G 1G7; e-mail: [email protected].
negative effects pertain to emotional and behavioral functioning, social competence, school achievement, cognitive functioning, psychopathology, and general health. Although some of these effects have been replicated across studies and generally fit with theo- retical and clinical expectations, there are a number of methodological issues that cloud their interpretation.
Researchers acknowledge that exposure to do- mestic violence is a nonspecific risk factor for devel- opmental harm, typifying the process of multifinality of development (Sameroff, 2000). That is, such expo- sure is part of a group of harm-producing contextual factors (such as child abuse, harsh parenting practices, and other forms of trauma and violence) that interfere with normal development and lead to unpredictable, but generally negative, outcomes in the short- and long-term. Embedded in the literature is the further assumption that exposure to domestic violence cre- ates a negative impact on children’s emotional and behavioral adjustment over and above other coexist- ing factors (i.e., it is not merely a confound or a cor- relate). Saunders (2003) underscores this latter point
171
1096-4037/03/0900-0171/0 C© 2003 Plenum Publishing Corporation
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
172 Wolfe, Crooks, Lee, McIntyre-Smith, and Jaffe
by reporting how different types of violence often co- occur in children’s lives and, coupled with the comor- bidity of problem outcomes, investigations of any sin- gle type of violence face considerable challenges.
Although recognizing that exposure to domestic violence is often harmful, researchers have cautioned that the heterogeneity of the population, variability in findings across studies, and many other method- ological limitations pose considerable obstacles to implying a cause and effect relationship (Edleson, 1999; Fantuzzo & Lindquist, 1989). Given the num- ber of methodological concerns that have been noted in these reviews, it is important to examine the overall empirical findings (including possible moderators of the impact of domestic violence on children) with a cautious lens. Significant remaining barriers include, for example, sampling concerns (i.e., generalization from shelter samples), reliance on mother-only rat- ings of children’s behavior, defining exposure to vio- lence, and separating the impact of child abuse from indirect exposure to domestic violence. As well, the literature suffers from a lack of theoretical clarity and guidance, which seems especially warranted given the complexity of the issues involved.
A DEVELOPMENTAL PSYCHOPATHOLOGY FRAMEWORK
Developmental psychopathology provides a use- ful framework for organizing the study of children’s exposure to domestic violence. This framework em- phasizes the role of developmental processes, the im- portance of context, and the influence of multiple and interacting events in shaping adaptive as well as mal- adaptive development (Rutter & Sroufe, 2000). More- over, the importance and complexity of family, social, and cultural factors are acknowledged in predicting and understanding developmental changes and ab- normal outcomes, and single-variable causes are held to greater scrutiny. This framework, therefore, con- siders how children adapt to harsh events in their daily surroundings, such as direct and indirect forms of violence, at the expense of important regulatory processes, which compromises their ongoing devel- opment. For example, episodes of violence and abuse between family members may prompt efforts on the part of the child to accommodate to such events and form a hypervigilant, insecure approach to relation- ships, often marked by strong emotions (e.g., frustra- tion, disappointment, hostility, fear). From this per- spective, children’s varied emotional and behavioral
problems associated with exposure to domestic vio- lence are understandable in that they represent efforts to adapt to a maladaptive situation.
Developmental psychopathology further raises the importance of a multidimensional, interactive ap- proach to the study of children’s development, in contrast to static comparisons of one-to-one relation- ships. This view, shared by other perspectives as well, argues that there is rarely a direct causal pathway leading to a particular outcome; instead, there are ongoing interactions between protective and vulner- ability factors within the child, between the child and his or her surroundings, and among particular risk factors. These factors are processes rather than ab- solutes, since the same event or condition can func- tion as either a protective or a vulnerability factor depending on the overall context in which it occurs. Acute, stressful situations as well as chronic adver- sity put children’s successful development at risk, yet these critical variables have rarely been controlled or examined in relation to domestic violence exposure. Finally, this perspective highlights the need to exam- ine children’s normal and abnormal development in relation to meaningful moderators such as sex, age, and type of outcome.
Whereas studies of children’s exposure to do- mestic violence have acknowledged the importance of potential moderators that may affect developmen- tal pathways, there has been only limited success at understanding their significance or roles. Shelter res- idence of the sample, reliance on maternal report of children’s behavior, and efforts to control for direct and indirect exposure to violence are commonly men- tioned as potential confounds or critical independent variables. For example, because children exposed to violence may present with similar difficulties to those who are direct victims of abuse, it is difficult to de- termine the degree to which behavioral outcomes are attributable to one or the other (Saunders, 2003). Sec- ond, there may be important differences in the char- acteristics of the families and the symptoms of chil- dren in these two groups (i.e., spouse-abusive versus spouse- and child-abusive), which get obscured when children exposed to domestic violence are grouped together regardless of child abuse status (Shipman, Rossman, & West, 1999). As a final consideration, the presence of child abuse raises the issue of mul- tiple risk factors or stressors, and is an important re- minder of contextually valid research (see Margolin & Gordis, 2000). Children’s age and sex are also noted as being especially relevant in interpreting the effects of exposure to domestic violence although, like the
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 173
contextual factors noted above, practical limitations in sample selection often preclude any separate anal- yses of these variables.
In summary, there is a wide degree of method variance in the research conducted on children ex- posed to domestic violence. These differences make it difficult to compare across studies due to different def- initions, samples, and methodology, and to draw firm conclusions. Nonetheless, it is important to look for common elements in these findings and examine theo- retically relevant variables. Therefore, the main objec- tive of this meta-analysis was to summarize the empir- ical findings of research evaluating consequences of domestic violence exposure on children. In addition, attempts were made to identify moderators that might lend more precision to the wide range of methodolo- gies used in the area. The limits of the current lit- erature are identified, in addition to suggestions for advancing research in this area. We approached the task of meta-analysis with these limitations and needs firmly in mind, and therefore chose to use a conser- vative approach to examining overall effect sizes and potential moderators. Thus, the current meta-analysis analyzes a smaller set of studies that more accurately assess exposure to domestic violence, while keep- ing in mind the potential methodological limitations described above.
METHOD
Meta-analytic techniques facilitate the synthesis of a large number of studies by distilling the empiri- cal results to interpretable averages, thus potentially identifying emerging themes across studies. One ap- proach is to include every study that has been con- ducted in an area, with the assumption that underly- ing “truths” will be identified, and that the advantages gained by a larger sample size outweigh the disadvan- tages of potentially faulty design or logic in any one study. This approach, used in a recent meta-analysis on exposure to domestic violence, facilitates the inclu- sion of a large number of studies (Kitzmann, Gaylord, Holt, & Kenny, 2003). A competing model is the one espoused by the Campbell Collaboration (n.d.), an international nonprofit organization that aims to help researchers make well-informed decisions about the effects of interventions in the social, behavioral, and educational arenas. The goal of this collaboration is to stimulate the empirical methodological research required to improve the validity, relevance, and preci- sion of systematic reviews and the randomized trials
and nonrandomized trials on which they are based. Rather than including all available studies in a meta- analysis or review, this approach advocates a theoret- ically driven approach that relies on drawing conclu- sion from studies that meet more stringent criteria.
The starting point for any meta-analysis is to con- ceptually define the parameters of study with respect to independent and dependent variables. Studies of the effects of children’s exposure to domestic violence have used wide-ranging parameters and various pop- ulations, including parental conflict and divorce, ret- rospective studies of exposure experienced by adults as children, simulated conflict studies, and studies of children exposed to domestic violence in their homes. In contrast to Kitzmann et al. (2003), the current anal- ysis incorporated only those studies that pertained to children exposed to domestic violence. This decision was based on the knowledge that there are so many existing sources of error and variability in these stud- ies (shelter status, clinical versus nonclinical samples, etc.) that a more narrowly defined independent vari- able increases the interpretability of the results.
PsycInfo and the National Clearinghouse on Child Abuse and Neglect databases were searched using the terms: [(domestic or interparental or mari- tal) AND (exposure or witness*) AND (violence or conflict or abuse or battered) AND (child or children or youth)]. A manual search of the references of re- view articles was also conducted to supplement the electronic searches. Approximately 400 journal arti- cles were initially found and evaluated with respect to the inclusion criteria (below), and of those 40 met our criteria and were included in the present analyses. These 40 articles actually represent 41 studies as one article had two separate studies (Jouriles, Norwood, McDonald, Vincent, & Mahoney, 1996). The final 41 studies had been published in peer-reviewed journals and included behavioral and/or emotional outcome measures of children’s adjustment. See Table I for de- scriptive information.
Inclusion and Exclusion Criteria
To calculate effect sizes, only studies that pub- lished means and standard deviations for at least two groups (i.e., a group of children exposed to domestic violence and an appropriate nonexposed comparison group) or correlations within a target group were an- alyzed. A small number of methodologically sound studies were excluded because they provided statis- tics that were not applicable with meta-analysis (e.g., multiple regression coefficients).
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Ta bl
e I.
D es
cr ip
ti on
s of
St ud
ie s
U se
d in
M et
a- A
na ly
si s
P ar
ti ci
pa nt
gr ou
ps (r
el ev
an tt
o m
et a-
an al
ys is
) O
ut co
m es
m ea
su re
d To
ta l
B oy
s G
ir ls
A ge
ra ng
e Ta
rg et
Ta rg
et C
om pa
ri so
n C
it at
io n
(N )
(N )
(N )
(y ea
rs )
E th
ni ci
ti es
G ro
up 1
G ro
up 2
G ro
up A
bu se
d/ W
it ne
ss Sh
el te
r R
at er
sa In
t. E
xt .
P T
SD O
th er
C hr
is to
po ul
os et
al .
(1 98
7) b
67 —
— 5–
13 79
% C
au ca
si an
; 21
% A
fr ic
an A
m er
ic an
Sh el
te r
(n =
37 )
C om
m un
it y
C on
tr ol
(n =
30 )
Y es
O ne
× ×
P er
ce iv
ed co
m pe
te nc
e
C oy
ne ,B
ar re
tt ,&
D uf
fy (2
00 0)
18 18
0 7–
11 N
ot sp
ec ifi
ed W
it ne
ss es
(n =
10 )
C om
m un
it y
co nt
ro l(
n =
8) N
o O
ne ×
× C
P IC
th re
at ;
B la
m e
D uR
an t,
C ad
en he
ad ,
P en
de rg
ra st
, Sl
av en
s, &
L in
de r
(1 99
4)
22 5
99 12
6 11
–1 9
10 0%
A fr
ic an
A m
er ic
an C
om m
un it
y (i
nn er
-c it
y; n =
22 5)
N o
O ne
× A
do le
sc en
tu se
of vi
ol en
ce
E l-
Sh ei
kh &
H ar
ge r
(2 00
1) 86
46 43
8– 11
81 %
C au
ca si
an ;8
% A
fr ic
an A
m er
ic an
;2 %
H is
pa ni
c; 2%
A si
an ;7
% ot
he r
C om
m un
it y
(n =
86 )
N o
M ul
ti pl
e ×
× C
P IC
th re
at ;
B la
m e
Fa nt
uz zo
et al
.( 19
91 )
77 41
36 3–
6 59
% C
au ca
si an
;5 %
A fr
ic an
A m
er ic
an ;2
9% H
is pa
ni c;
4% A
bo ri
gi na
l; 4%
A si
an ;2
% m
ix ed
Sh el
te r
(V er
ba l/P
hy si
ca l;
n =
23 )
H om
e (V
er -
ba l/P
hy si
ca l;
n =
27 )
C on
tr ol
(n =
27 )
P ar
ti al
gr ou
ps O
ne ×
× C
B C
L so
ci al
co m
p.
G ra
ha m
-B er
m an
n (1
99 6)
12 1
59 62
7– 12
40 %
C au
ca si
an ;
60 %
un sp
ec ifi
ed (m
aj or
it y
A fr
ic an
A m
er ic
an )
Sh el
te r
(n =
60 )
C om
m un
it y
co nt
ro l
(n =
61 )
Y es
O ne
× ×
G ry
ch ,F
in ch
am ,
Jo ur
ile s,
& M
cD on
al d
(2 00
0)
46 4
23 5
22 9
10 –1
4 W
it ne
ss gr
ou p:
33 %
C au
ca si
an ;3
2% A
fr ic
an A
m er
ic an
;3 2%
H is
pa ni
c; 1%
ot he
r
Sh el
te r
(n =
14 5)
C om
m un
it y
co nt
ro l
(n =
31 9)
Y es
O ne
c ×
C P
IC Se
lf -b
la m
e; T
hr ea
t
H er
sh or
n &
R os
en ba
um (1
98 5)
32 32
0 N
ot sp
ec ifi
ed (M
1 =
8. 5;
M 2 =
9. 6)
N ot
sp ec
ifi ed
W it
ne ss
(n =
15 )
C om
m un
it y
co nt
ro l
(n =
18 )
N o
O ne
× ×
H ol
de n
& R
it ch
ie (1
99 1)
74 35
39 1–
8 N
ot sp
ec ifi
ed (m
aj or
it y
C au
ca si
an )
Sh el
te r
(n =
37 )
C om
m un
it y
co nt
ro l
(n =
37 )
Y es
O ne
× ×
Te m
pe ra
m en
t
H ug
he s
(1 98
8) 18
0 86
94 3–
12 N
ot sp
ec ifi
ed Sh
el te
r (n =
40 )
C om
m un
it y
co nt
ro l
(n =
83 )
A bu
se d/
W it
ne ss
(n =
55 )
Y es
M ul
ti pl
e ×
×
H ug
he s,
P ar
ki ns
on ,&
V ar
go (1
98 9)
15 0
ns ns
4– 12
N ot
sp ec
ifi ed
Sh el
te r
(n =
44 )
C om
m un
it y
co nt
ro l
(n =
66 )
A bu
se d/
W it
ne ss
(n =
40 )
Y es
M ul
ti pl
e ×
× C
B C
L so
ci al
co m
p.
In go
ld sb
y, Sh
aw ,
O w
ns ,&
W ils
lo w
(1 99
9)
12 9
12 9
0 T
2: 3.
5 ye
ar s;
T 3:
5 ye
ar s
77 %
C au
ca si
an ;
20 %
A fr
ic an
A m
er ic
an ;3
% ot
he r
C om
m un
it y
lo ng
it ud
in al
(n =
12 9)
N o
M ul
ti pl
e ×
×
Ja ff
e, W
ol fe
,W ils
on ,&
Z ak
(1 98
6) d
47 47
0 4–
16 N
ot sp
ec ifi
ed Sh
el te
r (n =
32 )
C om
m un
it y
co nt
ro l
(n =
15 )
Y es
O ne
× ×
C B
C L
so ci
al co
m p.
Jo ur
ile s,
B ar
lin g,
& O
’L ea
ry (1
98 7)
45 22
23 5–
13 N
ot sp
ec ifi
ed W
it ne
ss (n =
45 )
N o
M ul
ti pl
e ×
× B
P C
m ot
or ex
ce ss
; P
sy ch
ot ic
Jo ur
ile s,
M ur
ph y,
e
O ’L
ea ry
(1 98
9) 87
41 46
5– 12
N ot
sp ec
ifi ed
C lin
ic al
m ar
it al
(n =
87 )
N o
M ul
ti pl
e ×
× B
P C
in ad
eq ua
cy
Jo ur
ile s,
et al
. (1
99 6a
)e ,f
55 23
32 5–
12 95
% C
au ca
si an
;5 %
un sp
ec ifi
ed C
lin ic
al m
ar it
al (n =
55 )
N o
M ul
ti pl
e ×
×
174
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Jo ur
ile s
et al
.( 19
96 b)
f 19
9 10
6 93
5– 12
39 %
C au
ca si
an ;
37 %
H is
pa ni
c; 24
% A
fr ic
an A
m er
ic an
Sh el
te r
(n =
19 9)
Y es
O ne
× ×
Se lf
-e st
ee m
Jo ur
ile s,
Sp ill
er ,
St ep
he ns
, M
cD on
al d,
& Sw
an k
(2 00
0)
15 4
83 71
8– 12
40 %
C au
ca si
an ;
32 %
A fr
ic an
A m
er ic
an ;2
7% H
is pa
ni c;
1% ot
he r
Sh el
te r
(n =
15 4)
Y es
M ul
ti pl
e ×
× C
P IC
th re
at
K em
pt on
, M
cC om
bs -T
ho m
as ,
& Fo
re ha
nd (1
98 9)
48 29
19 11
–1 5
N ot
sp ec
ifi ed
C om
m un
it y
(n =
48 )
N o
M ul
ti pl
e ×
× Te
ac he
r- ra
te d
co m
pe te
nc e
K er
ig (1
99 8)
17 4
88 86
7– 11
85 %
C au
ca si
an ;
10 %
A si
an ;1
% ea
ch A
fr ic
an A
m er
ic an
, H
is pa
ni c,
& A
bo ri
gi na
l; 2%
ot he
r
C om
m un
it y
(n =
17 4)
N o
M ul
ti pl
e ×
× C
P IC
th re
at ;
Se lf
-b la
m e
K ilp
at ri
ck et
al .(
19 97
) 35
18 17
6– 12
N ot
sp ec
ifi ed
W it
ne ss
(n =
20 )
C om
m un
it y
co nt
ro l
(n =
15 )
N o
M ul
ti pl
e ×
K ol
bo (1
99 6)
60 30
30 8–
11 73
% C
au ca
si an
;7 %
A fr
ic an
A m
er ic
an ;5
% H
is pa
ni c;
3% A
bo ri
gi na
l; 12
% ot
he r
W it
ne ss
(n =
60 )
N o
O ne
× ×
L ev
en do
sk y,
H ut
h- B
oc ks
,S em
el ,
& Sh
ap ir
o (2
00 2)
62 25
37 3–
5 42
% A
fr ic
an A
m er
ic an
;2 4%
C au
ca si
an ;1
9% m
ix ed
;1 5%
ot he
r
C om
m un
it y
(n =
62 )
N o
O ne
×
L it
ro w
ni k
et al
.( 20
03 )
68 2
34 1
34 1
6 ye
ar ol
ds on
ly 52
% A
fr ic
an A
m er
ic an
;3 6%
C au
ca si
an ;1
2% ot
he r
C om
m un
it y
lo ng
it ud
in al
(n =
68 2)
N o
O ne
× ×
M ar
ti n
& C
le m
en ts
(2 00
2) 48
21 27
4 ye
ar ol
ds on
ly 96
% C
au ca
si an
4% un
sp ec
ifi ed
C om
m un
it y
(n =
48 )
N o
M ul
ti pl
e ×
×
M at
hi as
,M er
ti n,
& M
ur ra
y (1
99 5)
g 44
— —
6– 12
N ot
sp ec
ifi ed
Sh el
te r
(n =
22 )
C om
m un
it y
co nt
ro l
(n =
22 )
Y es
(r ec
en t)
M ul
ti pl
e ×
× A
da pt
iv e
be ha
vi or
s
M cC
lo sk
ey ,
So ut
hw ic
k, Fe
rn an
de z-
E sq
ue r,
& L
oc ke
(1 99
5)
48 25
23 5–
12 10
0% H
is pa
ni c
(M ex
ic an
A m
er ic
an )
W it
ne ss
(n =
24 )
C om
m un
it y
co nt
ro l
(n =
24 )
N o
M ul
ti pl
e ×
× ×
M cD
on al
d, Jo
ur ile
s, N
or w
oo d,
W ar
e, &
E ze
ll (2
00 0)
90 70
20 4–
7 79
% C
au ca
si an
; 10
% A
fr ic
an A
m er
ic an
;1 0%
H is
pa ni
c; 1%
ot he
r
C lin
ic al
ch ild
(n =
43 )
C lin
ic al
co nt
ro l
(n =
47 )
N o
M ul
ti pl
e ×
×
M cG
ee ,W
ol fe
,& W
ils on
(1 99
7) 16
0 70
90 11
–1 7
96 %
C au
ca si
an ;4
% un
sp ec
ifi ed
(A bo
ri gi
na l&
A fr
ic an
C an
ad ia
n)
C P
S ag
en cy
(n =
16 0)
N o
M ul
ti pl
e ×
×
175
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Ta bl
e I.
C on
ti nu
ed
P ar
ti ci
pa nt
gr ou
ps (r
el ev
an tt
o m
et a-
an al
ys is
) O
ut co
m es
m ea
su re
d To
ta l
B oy
s G
ir ls
A ge
ra ng
e Ta
rg et
Ta rg
et C
om pa
ri so
n C
it at
io n
(N )
(N )
(N )
(y ea
rs )
E th
ni ci
ti es
G ro
up 1
G ro
up 2
G ro
up A
bu se
d/ W
it ne
ss Sh
el te
r R
at er
s In
t. E
xt .
P T
SD O
th er
M ul
le r
et al
.( 20
00 )
65 25
40 13
–1 7
63 %
C au
ca si
an ;
22 %
A fr
ic an
A m
er ic
an ;1
5% H
is pa
ni c
C lin
ic al
ch ild
(n =
65 )
N o
O ne
× ×
×
O ’B
ri en
et al
.( 19
97 )
43 23
20 8–
12 60
% C
au ca
si an
; 14
% A
fr ic
an A
m er
ic an
;1 2%
H is
pa ni
c; 7%
A si
an ;7
% ot
he r
C om
m un
it y
sa m
pl e
(n =
43 )
N o
M ul
ti pl
e ×
× Se
lf -w
or th
O ’K
ee fe
(1 99
5) 18
4 93
91 7–
13 42
% C
au ca
si an
; 37
% H
is pa
ni c;
21 %
A fr
ic an
A m
er ic
an
Sh el
te r
(n =
12 0)
A bu
se d/
W it
ne ss
(n =
64 )
Y es
O ne
× ×
O so
fs ky
,W ew
er s,
H an
n, &
Fi ck
(1 99
3)
53 ns
ns 9–
12 10
0% A
fr ic
an A
m er
ic an
C om
m un
it y
(n =
53 )
N o
O ne
× ×
Po rt
er &
O ’L
ea ry
(1 98
0) 64
37 27
5– 16
N ot
sp ec
ifi ed
C lin
ic al
ch ild
(n =
64 )
N o
O ne
× ×
R og
er s
& H
ol m
be ck
(1 99
7) 80
28 52
11 –1
5 36
% C
au ca
si an
; 25
% A
fr ic
an A
m er
ic an
;2 3%
H is
pa ni
c; 3%
A si
an ;8
% E
as t
In di
an ;5
% m
ix ed
C om
m un
it y
(s ch
oo l;
n =
80 )
N o
O ne
× ×
Sp ac
ca re
lli ,S
an dl
er ,&
R oo
sa (1
99 4)
29 1
14 4
14 7
9– 12
50 %
C au
ca si
an ;
24 %
H is
pa ni
c; 14
% A
fr ic
an A
m er
ic an
;4 %
A bo
ri gi
na l;
8% ot
he r
C om
m un
it y
(i nn
er -c
it y;
n =
29 1)
N o
M ul
ti pl
e ×
× Se
lf -e
st ee
m
St er
nb er
g et
al .(
19 93
) 77
43 34
8– 12
10 0%
C au
ca si
an (I
sr ae
l) W
it ne
ss (n =
16 )
C om
m un
it y
co nt
ro l
(n =
31 )
A bu
se d/
W it
ne ss
(n =
30 )
N o
M ul
ti pl
e ×
×
Ta ng
(1 99
7) 39
17 22
6– 13
10 0%
A si
an (C
hi ne
se )
Sh el
te r
(n =
21 )
C om
m un
it y
co nt
ro l
(n =
18 )
Y es
M ul
ti pl
e ×
×
Ta nn
en ba
um et
al .
(1 99
2) 27
0 13
5 13
5 11
–1 5
10 0%
C au
ca si
an C
om m
un it
y (n =
27 0)
N o
O ne
× ×
W ol
fe ,J
af fe
,W ils
on ,&
Z ak
(1 98
5) 19
8 98
10 2
4– 16
N ot
sp ec
ifi ed
Sh el
te r
(n =
10 2)
C om
m un
it y
co nt
ro l
(n =
96 )
Y es
O ne
× ×
W ol
fe ,Z
ak ,W
ils on
,& Ja
ff e
(1 98
6) 63
35 28
4– 13
N ot
sp ec
ifi ed
Sh el
te r
(c ur
re nt
; n =
17 )
Sh el
te r
(f or
m er
; n =
23 )
C om
m un
it y
co nt
ro l
(n =
23 )
P ar
ti al
gr ou
ps O
ne ×
×
To ta
l 50
88 24
97 h
22 82
h 39
37 4
M ea
n 6.
6– 12
.0 ye
ar s
59 %
C au
ca si
an ;
25 %
A fr
ic an
A m
er ic
an ;1
9% H
is pa
ni c;
11 %
A si
an ;7
% ot
he r
a N
o. of
ra te
rs fo
r ch
ild ou
tc om
e m
ea su
re s
in cl
ud ed
in th
e m
et a-
an al
ys is
. b
M ea
ns fr
om na
rr ow
ed sa
m pl
e (i
n w
hi ch
Ss w
it ho
ut D
V w
er e
ta ke
n ou
to ft
he D
V gr
ou p,
an d
th os
e Ss
w it
h D
V w
er e
ta ke
n ou
to ft
he co
m pa
ri so
n gr
ou p)
. c O
ut co
m e
m ea
su re
s w
er e
ob ta
in ed
fr om
m ul
ti pl
e so
ur ce
s, bu
to nl
y ch
ild re
po rt
s ha
d in
fo rm
at io
n fo
r bo
th th
e w
it ne
ss an
d co
m pa
ri so
n gr
ou ps
. d
A ls
o ha
s an
“a bu
se on
ly ”
gr ou
p no
ti nc
lu de
d in
th e
an al
ys is
. e O
ve rl
ap pi
ng Ss
. f Jo
ur ile
s an
d co
lle ag
ue s
19 96
a an
d 19
96 b
re pr
es en
td if
fe re
nt st
ud ie
s (a
nd sa
m pl
es )
w it
hi n
a si
ng le
pa pe
r. g
O nl
y P
ha se
II of
M at
hi as
et al
.( 19
95 )
w as
in cl
ud ed
in th
e m
et a-
an al
ys is
.T he
P ha
se II
sa m
pl e
in cl
ud es
a su
bs et
of pa
rt ic
ip an
ts fr
om P
ha se
I; ho
w ev
er ,d
em og
ra ph
ic in
fo rm
at io
n w
as on
ly pr
ov id
ed fo
r th
e la
rg er
P ha
se I
sa m
pl e.
h E
st im
at es
du e
to m
is si
ng in
fo rm
at io
n on
se x.
176
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 177
Studies that only involved comparison groups of abused children, abused/witnesses, or witnesses of a lesser severity of domestic violence were excluded be- cause they overlap with the target group (i.e., both groups of children had been exposed to some form or degree of interparental violence; Jouriles et al., 1998). Multiple articles that relied on the same sample of children were not entered; in these cases, the article with the most comprehensive results that met the in- clusion criteria was used. In addition, studies were excluded that reported two groups but used a stan- dardized norm group as their comparison (e.g., studies that reported the results from the CBCL standardiza- tion group as their control group). Studies focusing primarily on interparental conflict (but not violence) were excluded, as were studies that used children’s reactions to a simulated conflict paradigm.
Effect sizes were calculated in the present study for behavioral (e.g., externalizing, conduct problems), emotional (e.g., internalizing, depression, anxiety), and Posttraumatic Stress Disorder (PTSD) outcome measures. An overall effect size was calculated for each study by taking the average of the behav- ioral, emotional, PTSD, and social problem effects (weighted by sample size, where relevant). Studies that examined other constructs (e.g., cognitive attri- butions, emotional encoding, perceptions, reactions to conflict vignettes, self-esteem) and did not include measures of the aforementioned outcomes were ex- cluded. Retrospective studies with adult participants recalling childhood experiences were also excluded. A recent twin study was excluded because the sam- ple size (N = 1, 103 twin pairs) was so much larger than any of the other studies (Jaffee, Moffitt, Caspi, Taylor, & Arsenault, 2002). Because a meta-analysis approach weights the effect sizes by relative sample size, this study would have been disproportionately responsible for the overall outcome. The effect size for this study was .17 for the total sample of twin pairs (N = 2, 206).
Definitions of Confounding Variables
Despite the narrower focus of this meta-analysis, considerable variability remained with respect to the determination of key variables. One construct that defies precise definition and measurement is the na- ture and extent of exposure to domestic violence. Al- though a wide range of variation is recognized in the types, severity, and chronicity of violence experienced by women in intimate relationships (Holtzworth-
Munroe & Stuart, 1994), the heterogeneity of these experiences tends to be overlooked when the focus shifts to the children of these women (see Jouriles et al., 1996, 1998 for notable exceptions). The defi- nitions of interparental violence in the present anal- yses varied greatly, with a common definition being the endorsement of at least one physical incident in the past year (in contrast with a chronic history of se- vere battering). Previous research suggests that adults tend to vastly underestimate the extent to which their children are exposed (Jaffe, Wolfe, & Wilson, 1990; O’Brien, John, Margolin, & Erel, 1997). As well, there is a whole continuum of involvement for children, ranging from seeing the sequelae of violence or being passive observers, to attempts by children to physi- cally intervene or seek help. A recent telephone sur- vey of 114 battered women revealed that almost 25% of their children were reported to have been phys- ically involved in a battering incident, and over half verbally intervened while in the same room (Edleson, Mbilinyi, Beeman, & Hagemeister, 2003). To date, the literature on children exposed to domestic violence has overlooked this continuum of involvement, and tends to assume that exposure is a uniform experience. Source of information used to determine the presence of child abuse is likewise variably determined.
Table II summarizes how investigators of the studies included in this meta-analysis determined the key inclusion criteria of interparental violence, child exposure, and child abuse. To define the interparental violence group, 19 studies used maternal report only (which was often based on responses or direct ques- tioning from the Conflict Tactics scale; Straus, 1979); 15 studies relied on information obtained from multi- ple informants (which typically consisted of both par- ents, or mother and child in some cases); 3 assumed parental violence on the basis of shelter residence, and 4 on child report alone. Similarly, the majority of studies (13) assumed the presence of child exposure from maternal report or through direct questioning of the mother (16). Fewer studies (12) asked the children themselves about their exposure.
Very few studies controlled for the possible con- founding factor of child abuse, and many did not address the issue at all. Some studies assessed child abuse, yet did not utilize that information. Others controlled for child abuse in later analyses (e.g., hi- erarchical regression) using statistics that could not be incorporated into the meta-analysis (13). Rela- tively few studies (4) separated the domestic violence groups (i.e., witness versus abused/witness) or used child abuse as an exclusionary factor (3). For those
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Ta bl
e II
. C
ri te
ri a
fo r
D et
er m
in in
g In
te rp
ar en
ta lV
io le
nc e,
C hi
ld E
xp os
ur e
to V
io le
nc e,
an d
C hi
ld A
bu se
In te
rp ar
en ta
lv io
le nc
e C
hi ld
ex po
su re
C hi
ld ab
us e
M ot
he r
on ly
M ul
ti in
fo rm
an t
P ar
en ta
l M
at er
na ld
ir ec
t E
xa m
in ed
(n o
gr ou
p So
ur ce
of ch
ild Sh
el te
r (e
.g .,
C T
S, (e
.g .,
C T
S, C
hi ld
C T
S/ D
V re
po rt
(e .g
., m
od ifi
ed C
hi ld
Se pa
ra te
di ff
er en
ce s;
no tu
se d;
N ot
m al
tr ea
tm en
t as
su m
ed a
in te
rv ie
w )
in te
rv ie
w )
on ly
as su
m ed
C T
S, in
te rv
ie w
) re
po rt
gr ou
ps E
xc lu
de d
un us
ab le
st at
s fo
r m
et a)
ad dr
es se
d in
fo rm
at io
n
C hr
is to
po ul
os et
al .
(1 98
7) ×
× ×
N /A
C oy
ne et
al .(
20 00
) ×
× ×
In te
rv ie
w w
it h
m ot
he r
D uR
an te
ta l.
(1 99
4) ×
× ×
N /A
E l-
Sh ei
kh &
H ar
ge r
(2 00
1) ×
× ×
N /A
Fa nt
uz zo
et al
. (1
99 1)
× ×
× C
P S
re co
rd s
G ra
ha m
-B er
m an
n (1
99 6)
× ×
× N
/A
G ry
ch et
al .(
20 00
) ×
× ×
N /A
H er
sh or
n &
R os
en ba
um (1
98 5)
×b ×
× N
/A
H ol
de n
& R
it ch
ie (1
99 1)
× ×
× P
C -C
T S
(m ot
he r)
H ug
he s
(1 98
8) ×
× ×
M ot
he r
& sh
el te
r st
af fr
ep or
ts H
ug he
s et
al .(
19 89
) ×
× ×
M ot
he r
& sh
el te
r st
af fr
ep or
ts In
go ld
sb y
et al
. (1
99 9)
× ×c
× N
/A
Ja ff
e et
al .(
19 86
) ×
× ×
C P
S re
co rd
s Jo
ur ile
s et
al .(
19 87
) ×
× ×
P C
-C T
S (m
ot he
r re
po rt
s bo
th pa
re nt
s) Jo
ur ile
s et
al .(
19 89
) ×
× ×
N /A
Jo ur
ile s
et al
. (1
99 6a
) ×
× ×
N /A
Jo ur
ile s
et al
. (1
99 6b
) ×
× ×
N /A
Jo ur
ile s
et al
.( 20
00 )
× ×
× P
C -C
T S
(m ot
he r
& ch
ild re
po rt
s) K
em pt
on et
al .
(1 98
9) ×d
× ×
N /A
K er
ig (1
99 8)
× ×
× N
/A K
ilp at
ri ck
et al
. (1
99 7)
× ×
× M
ot he
r &
ch ild
qu es
ti on
na ir
es
178
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
K ol
bo (1
99 6)
× ×
× P
C -C
T S
(c ar
eg iv
er —
m aj
or it
y m
ot he
r) L
ev en
do sk
y et
al .
(2 00
2) ×
× ×
C hi
ld in
te rv
ie w
L it
ro w
ni k
et al
. (2
00 3)
× ×
× P
C -C
T S
(m ot
he r)
M ar
ti n
& C
le m
en ts
(2 00
2) ×
× ×
N /A
M at
hi as
et al
.( 19
95 )
× ×
× In
te rv
ie w
w it
h m
ot he
r M
cC lo
sk ey
et al
. (1
99 5)
× ×
× N
/A
M cD
on al
d et
al .
(2 00
0) ×
× ×
P C
-C T
S (m
ot he
r &
fa th
er )
M cG
ee et
al .(
19 97
) ×
× ×
A ll
Ss fr
om C
P S
ag en
cy /C
hi ld
re po
rt (R
P L
E )
M ul
le r
et al
.( 20
00 )
× ×
× C
hi ld
m ea
su re
(M yE
T V
) O
’B ri
en et
al .(
19 97
) ×
× ×
N /A
O ’K
ee fe
(1 99
5) ×
× ×
P C
-C T
S (m
ot he
r) O
so fs
ky et
al .(
19 93
) ×
× ×
N /A
Po rt
er &
O ’L
ea ry
(1 98
0) ×
× ×
N /A
R og
er s
& H
ol m
be ck
(1 99
7) ×
× ×
N /A
Sp ac
ca re
lli et
al .
(1 99
4) ×
× ×
So ci
al w
or ke
r re
po rt
s St
er nb
er g
et al
. (1
99 3)
× ×
× So
ci al
w or
ke r
(v al
id at
ed by
bo th
pa re
nt s
& ch
ild )
Ta ng
(1 99
7) ×
× ×
P C
-C T
S (m
ot he
r) Ta
nn en
ba um
et al
. (1
99 2)
× ×
× N
/A
W ol
fe et
al .(
19 85
) ×
× ×
P C
-C T
S (m
ot he
r) W
ol fe
et al
.( 19
86 )
× ×
× N
/A
To ta
ls 3
19 15
4 13
16 12
5 3
13 20
N ot
e. C
T S =
C on
fli ct
Ta ct
ic s
Sc al
e (S
tr au
s, 19
79 );
P C
-C T
S =
P ar
en t-
C hi
ld C
on fli
ct Ta
ct ic
s Sc
al e
(S tr
au s,
19 79
); R
P L
E =
R at
in gs
of P
as tL
if e
E ve
nt s
(M cG
ee ,1
99 0)
;M yE
T V =
M y
E xp
os ur
e to
V io
le nc
e (B
uk a
et al
., 19
96 ).
a St
ud y
m ay
ha ve
al so
ob ta
in ed
a D
V m
ea su
re (e
.g .,
C T
S) fo
r an
al ys
es ;h
ow ev
er ,g
ro up
in g
of D
V w
as de
te rm
in ed
by sh
el te
r st
at us
. b C
ri te
ri a
no ts
pe ci
fie d—
“h is
to ry
of do
m es
ti c
vi ol
en ce
.” c A
sk ed
w he
th er
ch ild
w it
ne ss
ed “d
is ag
re em
en ts
.” d B
ot h
m at
er na
la nd
pa te
rn al
C T
S re
sp on
se s
ob ta
in ed
.F or
th os
e st
ud ie
s w
hi ch
an al
yz ed
th es
e se
pa ra
te ly
,t he
m at
er na
lr ep
or tw
as en
te re
d.
179
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
180 Wolfe, Crooks, Lee, McIntyre-Smith, and Jaffe
studies that did assess for child abuse (21), the major- ity used only one informant (14), and only five queried the children themselves about abuse. One third of abuse assessments were conducted with the Parent– Child Conflict Tactics Scale (PC-CTS; Straus, 1979). Clearly, there is much heterogeneity between studies in terms of how domestic violence, exposure to such violence, and coexisting child abuse are addressed.
Data Coding and Analytical Approach
Means and standard deviations (for witnessing and comparison groups) and correlational data (for relating domestic violence to outcomes within witness groups) were used to generate effect sizes for each of the relevant outcomes. Effects were also coded for shelter sample, number of raters, and separate direct and indirect exposure to violence, to facilitate moderator analyses. Coding was independently veri- fied by a second (and in some cases third) rater. Dis- agreements or ambiguity regarding coding were re- solved through discussion among authors.
A total of 41 average effect sizes were included in this analysis (generated by combining measures of internalizing and externalizing difficulties, PTSD, and social problems for each study). The meta-analysis was conducted using methods outlined by Rosenthal (1995) and Wolf (1986). Effect sizes expressed as a standard difference score (d) were transformed to r scores to facilitate comparison of studies that re- ported either type of statistic. Next, a Fisher’s r to Z transformation was performed to standardize all of the effect sizes. Results are reported for a random effects analysis, which is appropriate because of the recognized variability within the sampling of studies (Borenstein & Rothstein, 1999).
Moderator Analysis
Studies that provided results separately for girls and boys, or those that included only males or females, were used to evaluate the sex of the child as a mod- erator. There were 10 studies that provided results for boys and girls separately, and another 4 that in- cluded boys only. To examine outcomes by develop- mental stage, age categories were chosen to represent preschool (3–6 years), school age (5–12 years), and adolescence (11 years +). Although these categories are somewhat arbitrary, they overlap to accommodate the inclusion of more studies in this moderator analy-
sis. Studies that had samples completely within one of those three categories were coded for developmental stage. Approximately 2/3 of the studies were coded (as shown in Table III), with the remaining ones not coded by age due to their large age span.
The various outcomes measured in the studies were divided into internalizing and externalizing cate- gories to examine outcome type as a moderator. Only studies that reported both types of measures were in- cluded, because using two effects (i.e., externalizing and internalizing) from some studies and only one effect (i.e., externalizing or internalizing) from other studies would result in overrepresentation of samples of children from studies that provided both types of outcome compared to those that did not. As a re- sult, none of the studies that measured PTSD was included,5 as they did not also include a measure of externalizing behavior. The measures used to deter- mine the externalizing and internalizing variables are provided in Table IV. If the Child Behavior Checklist (CBCL) Internalizing or Externalizing scores were available, subscales were not used. If the CBCL In- ternalizing score and an internalizing score from an- other measure (e.g., the Children’s Depression Inven- tory) were available, the average of these was used (weighted by subsample size, if relevant).
Although a developmental psychopathology framework should permit the evaluation of multiple dimensions in combination, too few studies provided results that would permit such analyses. As seen in Table III, only 7 of the studies that provided informa- tion by developmental stage also provide sex-specific results, and 12 provided information about both in- ternalizing and externalizing outcomes. Only four of the studies provided results that analyzed data with respect to three of these dimensions.
RESULTS
Results are presented in three sections: descrip- tive statistics of the characteristics of the studies; an overall meta-analysis across all outcomes; and mod- erator analyses.
5The average effect size for the three studies that measured PTSD symptomatology was r = .51(SD = .39). This large effect size should not be overinterpreted as the effects ranged from r = .16 (Muller et al., 2000) to r = .94 (Kilpatrick et al., 1997). Further- more, the total sample size of the three studies combined was only 162 participants. Clearly, the link between exposure to domestic violence and PTSD requires further study.
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 181
Table III. Studies Providing Sex and Outcome Results for a Particular Developmental Stage
Outcomes measured
Stage Citation N Overall effect Outcomes by sex Internalizing Externalizing PTSD
Preschool (3–6) Fantuzzo et al. (1991) 77 .51 × × Ingoldsby et al. (1999) 129 .25 × × Levendosky et al. (2002) 63 .42 × Litrownik et al. (2003) 583 .13 × × Martin & Clements (2002) 48 .33 × ×
Total (Preschool) 5 studies 900 0 4 4 1
School Age (5–12) Coyne et al. (2000) 18 .73 × × El-Sheikh & Harger (2001) 86 .11 × × Graham-Bermann (1996) 121 .14 × × Grych et al. (2000) 464 .12 × × Jouriles et al. (1996a) 55 .23 × × × Jouriles et al. (1996b) 199 .18 × × Jouriles et al. (2000) 154 .24 × × Kerig (1998) 174 .30 × × × Kilpatrick et al. (1997) 35 .94 Kolbo (1996)a 60 .26 × McCloskey et al. (1995)a 48 .33 × O’Brien et al. (1997) 43 .22 × × O’Keefe (1995) 185 .15 × × Osofsky et al. (1993) 53 .67 × × Spaccarelli et al. (1994) 131 .10 × × × Sternberg et al. (1993) 47 .54 × ×
Total (School) 16 studies 1,873 5 13 12 2
Adolescent (11–19) DuRant et al. (1994) 225 .29 × Kempton et al. (1989) 48 .21 × × McGee et al. (1997) 160 .07 × × × Muller et al. (2000) 65 .21 × × × Rogers & Holmbeck (1997) 80 .40 × × Tannenbaum et al. (1992) 224 −.20 × × ×
Total (Adolescent) 6 studies 802 2 6 5 1 aOverall CBCL problem score only (internalizing and externalizing not reported).
Table IV. Measures Grouped Into Internalizing and Externalizing Categories of Outcomes
Internalizing Externalizing
CBCL: Internalizing CBCL: Externalizing CBCL: Anxious/Depressed CBCL: Aggressive CBCL: Somatic CBCL: Attention Children’s Depression Inventory CBCL: Delinquent Revised Children’s Manifest BPC: Anxiety
Anxiety Scale (R)BPC: Personality (R)BPC: Conduct Problems BPC: Anxiety BPC: Aggression YSR: Internalizing ECBI: Intensity
ECBI: Frequency YSR: Externalizing Conners: Conduct Conners: Hyperactivity
Note. CBCL= Child Behavior Checklist; (R)BPC= (Revised) Be- havior Problem Checklist; YSR = Youth Self-Report; ECBI = Ey- berg Child Behavior Inventory.
Description of Samples
The summary of studies provided in Table I shows wide variability in participant characteristics across studies. Sample sizes reported for each study are minimum samples; that is, the reported N rep- resents the number of participants for which all of the relevant data were available. Age is reported as a range where available, as ranges were reported more often than means. In terms of shelter residence, 32% of the studies used shelter-only samples for their wit- ness group, 63% used nonshelter samples, and 5% used both (with separate groups). Of the 26 nonshel- ter samples, 50% were community samples, in which the number of families with interparental violence may have been relatively low, but scores on domes- tic violence measures were correlated with outcome measures. The proportion of studies with single versus multiple raters of child outcomes were similar: 51%
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
182 Wolfe, Crooks, Lee, McIntyre-Smith, and Jaffe
Table V. Shelter Residence, Number of Raters, Sex and Type of Child Outcome as Moderating Variables of Child Adjustment
Moderator n Zr SD 95% CI z-values z-score (diff)
Gender (10 studies) Boys 746 .11 .19 .04–.19 3.08 ns Girls 734 .09 .19 .02–.16 2.38
Type of outcome (31 studies) Internalizing 5148a .19 .12 .16–.21 13.59 ns Externalizing 5148 .21 .12 .18–.23 15.05
aSame samples, but different measures.
single (e.g., mother only, child only), and 49% multi- ple (e.g., mother and child, mother and teacher). Most studies had measures of child externalizing and inter- nalizing difficulties, except for those involving PTSD (which tended to exclude externalizing symptoms). Although the sex ratio was approximately equivalent overall, 10% of the studies included only boys. The majority of participants was Caucasian, although a fair number of other ethnicities was represented (see Table I).
Consequences of Exposure to Domestic Violence
Forty of the 41 effects indicated that exposure to domestic violence has a negative effect on children. Further support for an overall relationship between exposure to violence and negative outcomes was pro- vided by an aggregate weighted mean correlation of Zr = .28 (SD = .17; 95% CI = .21–.32), which is sig- nificant (Zc = 8.86, p < .001). A Zr = .28 effect cor- responds to a small effect size (Cohen, 1977). The one study that reported a reverse effect from what would be expected provided somewhat ambiguous results (Tannenbaum, Neighbors, & Forehand, 1992). In that study the overall correlation between exposure to vi- olence and problematic child outcomes was negative, but the unique contribution of exposure to violence in predicting poor outcomes (when other confounds were controlled) was positive. Thus, the unexpected result likely represents a complicated pattern of rela- tionships among domestic violence and other dynam- ics, and underscores the need for contextually relevant research.
Moderator Analyses
An analysis of heterogeneity was conducted to determine whether there was adequate dispersion of individual outcomes vis a vis the overall effect
to explore for possible moderators (Borenstein & Rothstein, 1999). Given evidence of significant het- erogeneity (χ2 = 188.49, df = 40, p < .001), a small number of variables was explored, with the results summarized in Table V. Results of a fixed effects anal- ysis are reported for the moderators as per convention (Borenstein & Rothstein, 1999).
Developmental Stage
When all 27 studies that had samples within a particular developmental stage were compared, the school aged children demonstrated the largest av- erage effect size (Zr = .23), followed by preschool- ers (Zr = .22) and adolescents (Zr = .11). The dif- ference across developmental stages was significant (Z= 8.76, p < .05). However, this analysis exempli- fies one of the problems that arises in using meta- analysis techniques with a small number of studies that have wide variability in methodology. The aver- age effect size for school aged children was strongly affected by one study (Kilpatrick, Litt, & Williams, 1997), in which the outcome of interest was PTSD. However, rather than using rates of diagnosis, results were reported with respect to a PTSD scale, and there was considerable dispersion in mean scores between the groups (with a corresponding effect size of Zr = .94). This outcome, therefore, may have been an arte- fact of the scale rather than the difference suggested by such an extreme effect size; alternatively, this may suggest that comparing mean scores for a syndrome such as PTSD may be misleading. Similarly, the effect size for adolescents may be artificially suppressed by the Tannenbaum et al.’s study (Tannenbaum et al., 1992), which generated a negative effect size.
When the developmental stage moderator anal- ysis was conducted without these two studies, signif- icant differences among developmental stages disap- peared. Adolescent (Zr = .23), preschool (Zr = .21), and school age samples (Zr = .21) showed similar
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 183
effect sizes. Rather than providing a basis for draw- ing conclusions about the effects of domestic violence at different stages of development, this example il- lustrates the lack of a clear result with respect to developmental stage, and underscores the variability across studies with respect to methodology. Because of the lack of stability for a solution concerning de- velopmental stage, this moderator is not presented in Table V.
Sex
Initially, studies that reported results for both sexes as well as those that only involved boys were in- cluded in the analysis, generating average effect sizes of Zr = .17 for boys and Zr = .09 for girls. When the four studies that only included boys were removed, this gap between boys and girls disappeared (Zr = .11 and .09, respectively). The substantial convergence between boys and girls achieved by removing the “boys only” studies suggests that the high effect sizes for those samples might be related to sample charac- teristics other than sex. The latter analysis is reported in Table V as the more conservative estimate of the two with respect to sex differences in the effects of exposure to domestic violence.
Type of Outcome
On the basis of the 31 studies that provided in- formation about both internalizing and externalizing adjustment problems, the moderator analysis with re- spect to type of outcome was not significant (Zr = .21 for externalizing, Zr = .19 for internalizing).
Comparison of Witnesses and Combined Witness/Victims
Although the intention at the outset of this pa- per was to examine the presence of direct victim- ization as a moderator of exposure to domestic vi- olence, meta-analysis was curtailed because of the availability of only four studies. The individual results of these studies are presented in Table VI for descrip- tive purposes only. Because there was significant vari- ability, internalizing and externalizing outcomes are presented separately for each study. Effect sizes are presented such that a positive effect corresponds to a finding that children who are both witnesses and
victims are functioning more poorly than those who only witness. The findings across these four studies suggest a small effect size for the difference between children who are combined witness/victims and those who are witnesses only. There is preliminary evidence that this difference is greater for externalizing behav- iors, although more studies are required to determine whether or not this trend is significant. Although the difference in outcomes between these two groups of children may be statistically nonsignificant, there is an issue of restricted range that is important to consider. That is, the comparison group in this case is children exposed to violence (and the target group has been exposed to direct and indirect violence). Thus, the ex- perience of direct victimization may add a small effect size in addition to the medium effect already present with respect to exposure to domestic violence.
DISCUSSION
The purpose of this article was to synthesize the collective literature on children exposed to domes- tic violence, with respect to negative emotional and behavioral outcomes. The answer to whether or not children exposed to violence experience more diffi- culties than their peers emerged as an unequivocal yes. When evaluated across all of the samples and outcomes, a small effect size was evident for expo- sure to domestic violence. In terms of translating an effect size of r = .28 into a more concrete concept, in the treatment literature an effect size of r = .30 would mean an increase in successful treatment rate from 35 to 65% (Wolf, 1986). Conversely, the variance equiv- alent to an effect of r = .30 could be interpreted as in- creasing the number of children exhibiting difficulties from 35 to 65%. Clearly, the statistical significance of exposure to domestic violence is matched by clinical significance. Furthermore, child abuse experiences (in addition to exposure) added a small increment in ef- fect size above and beyond exposure alone, although this finding is preliminary due to the limited number of studies.
The fundamental building blocks of developmen- tal theory (developmental stage, sex, and type of out- come) were examined as moderators. The lack of sig- nificant findings with these moderators, as well as the degree to which the results changed with minor alterations to the analyses, underscores the lack of sta- bility in the underlying data set. Another way of fram- ing these findings is to note that the disparity in sam- pling (i.e., shelter versus clinical versus community),
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
184 Wolfe, Crooks, Lee, McIntyre-Smith, and Jaffe
Table VI. Preliminary Effect Sizes Comparing Children Who Are Both Witnesses and Victims with Witnesses only
Internalizing Externalizing
Study N Zr Effect size Zr Effect size
Hughes (1988) 24 −.17 Smalla .22 Small Hughes et al. (1989) 56 .22 Small .28 Medium O’Keefe (1995) 185 .12 Small .18 Small Sternberg et al. (1993) 46 .22 Small .12 Small aEffect in reverse direction.
wide range in outcomes among these children, the wide method variance in measuring outcomes, and lack of a contextually sensitive approach produced greater variability across studies than that found be- tween sexes or across developmental stages. In other words, methodological variability and other unspeci- fied factors produced larger differences in effect sizes than did the selected moderators of age, sex, and type of outcome.
The current state of the literature on children exposed to domestic violence provides a solid foun- dation from which to move forward with more com- plicated hypotheses and analyses. In comparison to the literature on other forms of child maltreatment, the state of this literature is less developed. For exam- ple, early analyses of child sexual abuse sequelae have been followed by more detailed analyses that indicate that the impact of sexual abuse may vary according to many factors, such as severity of the abuse, age of onset, nature of perpetrator, patterns of disclosure, and support systems in place for the child (Oddone- Paolucci, Genuis, & Violato, 2001). The impact of ex- posure to violence is likewise a complex phenomenon that may be determined by a host of factors within the child’s environment, family, and individual character- istics. This experience is different from many other single traumatic events and requires complex, mul- tivariate models that examine the interplay between trauma and development.
Several reviews have pointed to the challenges that face abused women and their children in escaping from batterers. These challenges may include the dis- tress of repeated separations, ongoing violence dur- ing visitation, and prolonged child custody battles in court (Jaffe, Lemon, & Poisson, 2003; Jaffe, Poisson, & Cunningham, 2001). The field will require multisite studies that can capture these complexities with large enough data samples to examine all the variables of interest including changes at different stages of de- velopment. Furthermore, questions remain about the long-term effects in adult relationships that may not
be visible from traditional measures of child adjust- ment. There may be some specific effect on children’s knowledge and attitudes about violence in relation- ships and their sense of personal responsibility for domestic violence that is not captured by current measures.
The question of long-term adjustment versus short-term adaptation to crisis will only be answered with the use of longitudinal data. Furthermore, given the variability within the population of children ex- posed to domestic violence, large samples are re- quired to capture the full picture. There is also a pressing need to investigate a wider range of nega- tive outcomes. Although the initial intention was to include educational and cognitive outcomes in this meta-analysis, there were not enough studies to facil- itate this inclusion. There are sound theoretical rea- sons to expect exposure to violence to have an impact on cognition and learning and to further explore these links; however, it is misleading to present this as a well-documented finding at this time. Similarly, there is emerging evidence for the link between exposure to domestic violence and PTSD in children; however, the dearth of studies in this area makes it premature to offer anything other than tentative conclusions. Our preliminary analyses show that PTSD symp- toms appear to be one negative outcome, particularly for younger children. Lehmann’s study (Lehmann, 1997) of child witnesses also found significant PTSD symptomatology in over half of the sample, raising the possibility of an interaction between trauma and the developmental stage of the child at the time of exposure.
Future Research Directions/Promising Approaches
Although the concept of ecologically valid mod- els has become de rigeur in the developmental lit- erature, research on children exposed to domestic violence has tended to focus on these children in a
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 185
vacuum. The need to evaluate the contribution of exposure to domestic violence in tandem with other risk factors is paramount. Some research efforts have begun to look at the exposure to domestic violence within the context of exposure to community violence, (e.g., Muller, Goebel-Fabbri, Diamond, & Dinklage, 2000), which highlights the unique and shared char- acteristics of these related forms of trauma.
Recent studies have begun to address some of the fundamental methodological flaws in this liter- ature by employing longitudinal designs with large samples, multiple raters of child outcomes, and sophis- ticated multivariate techniques. One multisite, lon- gitudinal study, (LONGSCAN) reported the effects of exposure to violence at 3 years of age on exter- nalizing and internalizing behavior at age 6, and is continuing to follow these children (Litrownik, New- ton, Hunter, English, & Everson, 2003). Other lon- gitudinal studies (Ware et al., 2001) have compared maternal ratings of child outcomes during and fol- lowing shelter residence to examine whether moth- ers’ more negative ratings of their child’s behavior (compared to rating by teachers and shelter staff, diagnostic interviews by researchers, etc.) are a re- sult of their level of distress during shelter residence. Studies that have examined this issue of maternal rat- ings with a cross-sectional design (Morrel, Dubowitz, Kerr, & Black, 2003), have compared child outcomes across raters, and controlled for maternal victimiza- tion and depression in addition to maternal distress. Finally, multivariate techniques are being utilized to identify developmental profiles that children may ex- hibit following exposure to domestic violence. To il- lustrate, a recent cluster analysis of 228 children from shelters identified five clusters based on internalizing and externalizing outcomes, which could be distin- guished with respect to frequency of the children’s exposure to interparental violence, and child abuse (Grych, Jouriles, Swank, McDonald, & Norwood, 2000).
In sum, in contrast to many meta-analyses that summarize a large set of studies and provide con- clusive findings, results of the current meta-analysis should be considered as a preliminary springboard to further research on this topic. Important progress has been made in terms of isolating possible moderators of the impact of exposure to violence on children, but unanswered questions still remain. The field is beginning to move away from epidemiological studies emphasizing prevalence and extent of clinically signif- icant problems, towards a more refined developmen- tal focus on the interaction of risk and protective fac-
tors that mediate the impact of exposure to domestic violence.
REFERENCES
References marked with an asterisk indicate studies included in the meta-analysis.
Borenstein, M., & Rothstein, H. (1999). Comprehensive meta- analysis: A computer program for research synthesis. Englewood, NJ: Biostat.
Campbell Collaboration. (n.d.). What helps? What harms? Based on what evidence? Retrieved April 15, 2003, from http://www.campbellcollaboration.org/Fra/About.html
*Christopoulos, C., Cohn, D. A., Shaw, D. S., Joyce, S., Sullivan- Hanson, J., Kraft, S. P., & Emery, R. E. (1987). Children of abused women. I: Adjustment at time of shelter residence. Journal of Marriage and the Family, 49, 611–619.
Cohen, J. (1977). Statistical power analysis for the behavioral sci- ences (4th ed.). New York: Academic Press.
Conrad, M., & Hammen, C. (1989). Role of maternal depression in perceptions of child maladjustment. Journal of Clinical and Consulting Psychology, 57, 663–667.
*Coyne, J. J., Barrett, P. M., & Duffy, A. L. (2000). Threat vigilance in child witnesses of domestic violence: A pilot study utiliz- ing the ambiguous situations paradigm. Journal of Child and Family Studies, 9, 377–388.
*DuRant, R. H., Cadenhead, C., Pendergrast, R. A., Slavens, G., & Linder, C. W. (1994). Factors associated with the use of violence among urban black adolescents. American Journal of Public Health, 84, 612–617.
Edleson, J. L. (1995). Mothers and children: Understanding the links between woman battering and child abuse. Paper pre- sented at the National Institute of Justice Strategic Planning Workshop on Violence Against Women, Washington, DC. Re- trieved February 12, 2003, from www.mincava.umn.edu
Edleson, J. L. (1999). Children’s witnessing of adult domestic vio- lence. Journal of Interpersonal Violence, 14, 839–870.
Edleson, J. L., Mbilinyi, L. F., Beeman, S. K., & Hagemeister, A. K. (2003). How children are involved in domestic violence: Re- sults from a four city telephone survey. Journal of Interpersonal Violence, 18, 18–32.
*El-Sheikh, M., & Harger, J. (2001). Appraisals of marital conflict and children’s adjustment, health, and psychological reactivity. Developmental Psychology, 37, 875–885.
*Fantuzzo, J. W., DePaola, L. M., Lambert, L., Martino, T., Anderson, G., & Sutton, S. (1991). Effects of interparent vi- olence on the psychological adjustment and competencies of young children. Journal of Consulting and Clinical Psychology, 59, 258–265.
Fantuzzo, J. W., & Lindquist, C. U. (1989). The effects of observ- ing conjugal violence on children: A review and analysis of research methodology. Journal of Family Violence, 4, 77–94.
Fantuzzo, J. W., & Mohr, W. K. (1999). Prevalence and effects of child exposure to domestic violence. The Future of Children, 9, 21–32.
*Graham-Bermann, S. A. (1996). Family worries: Assessment of interpersonal anxiety in children from violent and non-violent families. Journal of Clinical Child Psychology, 25, 280–287.
*Grych, J. H., Fincham, F. D., Jouriles, E. N., & McDonald, R. (2000). Interparental conflict and child adjustment: Testing the mediational role of appraisals in the cognitive-contextual framework. Child Development, 71, 1648–1661.
Grych, J. H., Jouriles, E. N., Swank, P. R., McDonald, R., & Norwood, W. D. (2000). Patterns of adjustment among chil- dren of battered women. Journal of Consulting and Clinical Psychology, 68, 84–94.
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
186 Wolfe, Crooks, Lee, McIntyre-Smith, and Jaffe
*Hershorn, M., & Rosenbaum, A. (1985). Children of marital vio- lence: A closer look at the unintended victims. American Jour- nal of Orthopsychiatry, 55, 260–266.
*Holden, G. W., & Ritchie, K. L. (1991). Linking extreme marital discord, child rearing, and child behavior problems: Evidence from battered women. Child Development, 62, 311–327.
Holtzworth-Munroe, A., & Stuart, G. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116, 476–497.
*Hughes, H. M. (1988). Psychological and behavioral correlates of family violence in child witnesses and victims. American Journal of Orthopsychiatry, 58, 77–90.
*Hughes, H. M., Parkinson, D., & Vargo, M. (1989). Witness- ing spouse abuse and experiencing physical abuse: A ‘double whammy’? Journal of Family Violence, 4, 197–209.
*Ingoldsby, E. M., Shaw, D. S., Owens, E. B., & Winslow, E. B. (1999). A longitudinal study of interpersonal conflict, emo- tional and behavioral reactivity, and preschoolers’ adjustment problems among low-income families. Journal of Abnormal Child Psychology, 27, 343–356.
Jaffe, P. G., Lemon, N., & Poisson, S. (2003). Domestic violence and child custody disputes: Addressing the essential clinical and legal issues. Thousand Oaks, CA: Sage.
Jaffe, P. G., Poisson, S. E., & Cunningham, A. (2001). Domestic vio- lence and high-conflict divorce: Developing a new generation of research for children. In S. A. Graham-Bermann & J. L. Edleson (Eds.), Domestic violence in the lives of children: The future of research, intervention, and social policy. Washington, DC: American Psychological Association.
Jaffe, P., Wolfe, D., & Wilson, S. (1990). Children of battered women. Newbury Park, CA: Sage.
*Jaffe, P., Wolfe, D., Wilson, S., & Zak, L. (1986). Similarities in behavioral and social maladjustment among child victims and witnesses to family violence. American Journal of Orthopsy- chiatry, 56, 142–146.
Jaffee, S. R., Moffitt, T. E., Caspi, A., Taylor, A., & Arsenault, L. (2002). Influence of adult domestic violence on children’s internalizing and externalizing problems: An environmentally informative twin study. Journal of the American Academy of Child and Adolescent Psychiatry, 41, 1095–1103.
*Jouriles, E. N., Barling, J., & O’Leary, K. D. (1987). Predicting child behavior problems in martially violent families. Journal of Abnormal Child Psychology, 15, 165–173.
Jouriles, E. N., McDonald, R., Norwood, W. D., Ware, H. S., Spiller, L. C., & Swank, P. R. (1998). Knives, guns, and interparent violence: Relations with child behavior problems. Journal of Family Psychology, 12, 178–194.
*Jouriles, E. N., Murphy, C. M., & O’Leary, K. D. (1989). Inter- spousal aggression, marital discord, and child problems. Jour- nal of Consulting and Clinical Psychology, 57, 453–455.
*Jouriles, E. N., Norwood, W. D., McDonald, R., Vincent, J. P., & Mahoney, A. (1996). Physical violence and other forms of marital aggression: Links with children’s behavior problems. Journal of Family Psychology, 10, 223–234.
*Jouriles, E. N., Spiller, L. C., Stephens, N., McDonald, R., & Swank, P. (2000). Variability in adjustment of children of battered women: The role of child appraisals of interparent conflict. Cognitive Therapy and Research, 24, 233–249.
*Kempton, T., McCombs-Thomas, A., & Forehand, R. (1989). Di- mensions of interparental conflict and adolescent functioning. Journal of Family Violence, 4, 297–307.
*Kerig, P. K. (1998). Moderators and mediators of the effects of interparental conflict on children’s adjustment. Journal of Ab- normal Child Psychology, 26, 199–212.
*Kilpatrick, K. L., Litt, M., & Williams, M. (1997). Post-traumatic stress disorder in child witnesses to domestic violence. Amer- ican Journal of Orthopsychiatry, 67, 639–644.
Kitzmann, K. M., Gaylord, N. K., Holt, A. R., & Kenny, E. D. (2003). Child witnesses to domestic violence: A meta-analytic
review. Journal of Consulting and Clinical Psychology, 71, 339– 352.
*Kolbo, J. R. (1996). Risk and resilience among children exposed to family violence. Violence and Victims, 11, 113–128.
Lehmann, P. (1997). The development of posttraumatic stress dis- order (PTSD) in a sample of child witnesses to mother assault. Journal of Family Violence, 12, 241–257.
*Levendosky, A. A., Huth-Bocks, A. C., Semel, M. A., & Shapiro, D. L. (2002). Trauma symptoms in preschool-age children ex- posed to domestic violence. Journal of Interpersonal Violence, 17, 150–164.
*Litrownik, A., Newton, R., Hunter, W. M., English, D., & Everson, M. D. (2003). Exposure to family violence in young at-risk children: A longitudinal look at the effects of victimization and witnessed physical and psychological aggression. Journal of Family Violence, 18, 59–73.
Margolin, G., & Gordis, E. (2000). The effects of family and com- munity violence on children. Annual Reviews Psychology, 51, 445–479.
*Martin, S. E., & Clements, M. L. (2002). Young children’s respond- ing to interparental conflict: Associations with marital aggres- sion and child adjustment. Journal of Child and Family Studies, 11, 231–244.
*Mathias, J. L., Mertin, P., & Murray, A. (1995). The psychologi- cal functioning of children from backgrounds of domestic vi- olence. Australian Psychologist, 30, 47–56.
*McCloskey, L. A., Southwick, K., Fernandez-Esquer, M. E., & Locke, C. (1995). Psychological effects of political and do- mestic violence on Central American and Mexican immigrant mothers and children. Journal of Community Psychology, 23, 95–116.
*McDonald, R., Jouriles, E. N., Norwood, W., Ware, H. S., & Ezell, E. (2000). Husbands’ marital violence and the adjustment problems of clinic-referred children. Behavior Therapy, 31, 649–665.
*McGee, R. A., Wolfe, D. A., & Wilson, S. K. (1997). Multiple maltreatment experiences and adolescent behavior problems: Adolescents’ perspectives. Development and Psychopathol- ogy, 9, 131–149.
Morrel, T. M., Dubowitz, H., Kerr, M. A., & Black, M. M. (2003). The effect of maternal victimization on children: A cross-informant study. Journal of Family Violence, 18, 29– 41.
*Muller, R. T., Goebel-Fabbri, A. E., Diamond, T., & Dinklage, D. (2000). Social support and the relationship between fam- ily and community violence exposure and psychopathology among high risk adolescents. Child Abuse and Neglect, 24, 449– 464.
Oddone-Paolucci, E., Genuis, M. L., & Violato, C. (2001). A meta- analysis of the published research on the effects of child sexual abuse. Journal of Psychology, 135, 17–36.
*O’Brien, M., John, R. S., Margolin, G., & Erel, O. (1997). Reliabil- ity and diagnostic efficacy of parents’ reports regarding chil- dren’s exposure to marital aggression. Violence and Victims, 9, 45–52.
*O’Keefe, M. (1995). Predictors of child abuse in maritally violent families. Journal of Interpersonal Violence, 10, 3–25.
*Osofsky, J. D., Wewers, S., Hann, D. M., & Fick, A. C. (1993). Chronic community violence: What is happening to our chil- dren? Psychiatry, 56, 36–45.
*Porter, B., & O’Leary, K. D. (1980). Marital discord and childhood behavior problems. Journal of Abnormal Child Psychology, 8, 287–295.
*Rogers, M. J., & Holmbeck, G. N. (1997). Effects of interparental aggression on children’s adjustment: The moderating role of cognitive appraisal and coping. Journal of Family Psychology, 11, 125–130.
Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bulletin, 118, 183–192.
P1: GMX
Clinical Child and Family Psychology Review (CCFP) pp923-ccfp-469658 July 18, 2003 12:22 Style file version Nov. 07, 2000
Children Exposed to Domestic Violence 187
Rutter, M., & Sroufe, L. A. (2000). Developmental psychopathol- ogy: Concepts and challenges. Development and Psychopathol- ogy, 12, 265–296.
Sameroff, A. J. (2000). Developmental systems and psychopathol- ogy. Development and Psychopathology, 12, 297–312.
Saunders, B. E. (2003). Understanding children exposed to vio- lence: Toward an integration of overlapping fields. Journal of Interpersonal Violence, 18, 356–376.
Shipman, K., Rossman, B., & West, J. (1999). Co-occurrence of spousal violence and child abuse: Conceptual implications. Child Maltreatment, 4, 93–102.
*Spaccarelli, S., Sandler, I. N., & Roosa, M. (1994). History of spouse violence against mother: Correlated risks and unique effects in child mental health. Journal of Family Violence, 9, 79–98.
*Sternberg, K. J., Lamb, M. E., Greenbaum, C., Cicchetti, D., Dawud, S., Cortes, R. M., Krispin, O., & Lorey, F. (1993). Ef- fects of domestic violence on children’s behavior problems and depression. Developmental Psychology, 29, 44–52.
Straus, M. (1979). Measuring intrafamily conflict: The Conflict Tac- tics (CT) scales. Journal of Marriage and the Family, 41, 75–88.
*Tang, C. S. (1997). Psychological impact of wife abuse: Experiences of Chinese women and their children. Journal of Interpersonal Violence, 12, 466–478.
*Tannenbaum, L., Neighbors, B., & Forehand, R. (1992). The unique contribution of four maternal stressors to adoles- cent functioning. Journal of Early Adolescence, 12, 314– 325.
Ware, H. S., Jouriles, E. N., Spiller, L. C., McDonald, R., Swank, P. R., & Norwood, W. D. (2001). Conduct problems among children at battered women’s shelters: Prevalence and stabil- ity of maternal reports. Journal of Family Violence, 16, 291– 307.
Wolak, J., & Finkelhor, D. (1998). Children exposed to partner vio- lence. In J. Jasinski & L. M. Williams (Eds.), Partner violence: A comprehensive review of 20 years of research. (pp. 73–112). Thousand Oaks, CA: Sage.
Wolf, F. M. (1986). Meta-analysis: Quantitative methods for re- search synthesis. Quantitative Applications in the Social Sci- ences Series 07–059. Beverly Hills, CA: Sage.
*Wolfe, D. A., Jaffe, P. J., Wilson, S., & Zak, L. (1985). Children of battered women: Relation of child behavior to family vio- lence and maternal stress. Journal of Consulting and Clinical Psychology, 53, 657–665.
*Wolfe, D. A., Zak, L., Wilson, S., & Jaffe, P. (1986). Child witnesses to violence between parents: Critical issues in behavioral and social adjustment. Journal of Abnormal Child Psychology, 14, 95–104.
