You are on page 1of 16

Developmental Psychology Copyright 2007 by the American Psychological Association

2007, Vol. 43, No. 2, 438 – 453 0012-1649/07/$12.00 DOI: 10.1037/0012-1649.43.2.438

Child Negative Emotionality and Parenting From Infancy to Preschool:


A Meta-Analytic Review
Marja C. Paulussen-Hoogeboom, Geert Jan J. M. Stams, Jo M. A. Hermanns, and Thea T. D. Peetsma
University of Amsterdam

This meta-analytic review (k ⫽ 62 studies; N ⫽ 7,613 mother– child dyads) shows that effect sizes for
the association between child negative emotionality and parenting were generally small and were
moderated by sample and measurement characteristics. The association between more child negative
emotionality and less supportive parenting was relatively strong in lower socioeconomic status families,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

reversed in higher socioeconomic status families, and limited to studies with relatively high percentages
This document is copyrighted by the American Psychological Association or one of its allied publishers.

of participants from ethnic minorities and studies using parent report to assess negative emotionality.
Higher levels of child negative emotionality were associated with more restrictive control in samples with
less than 75% 1st-born children, as well as in infants and preschoolers, and in studies using parent report
or composite measures to assess both negative emotionality and restrictive parenting. Finally, more child
negative emotionality was associated with less inductive control.

Keywords: meta-analysis, negative emotionality, parenting, infancy, preschool

This meta-analytic study was aimed at investigating the strength is, much used and hence very influential (Prior, 1992; Wright
and nature of the association between children’s negative emotion- Guerin, Gottfried, & Thomas, 1997). Looking back, the NYLS can
ality in infancy and at preschool age and parenting behavior. be seen as the starting point for a new line of empirical research
Negative emotionality forms part of the temperament construct, studying the impact of child temperament on individual develop-
which refers to the behavioral style exhibited by infants or young ment. The NYLS also started a still-ongoing debate among re-
children in different contexts in response to a range of stimuli searchers about the conceptualization and measurement of temper-
(Zeanah & Fox, 2004). Temperament can be defined as constitu- ament. However, one broad aspect of temperament was agreed on
tionally based individual differences in emotional, motor, and relatively quickly, namely, the construct of negative emotionality.
attentional reactivity and self-regulation (Rothbart & Bates, 1998). Negative emotionality refers to irritability, negative mood,
The first major study of temperament was the New York Lon- (un)soothability, and high-intensity negative reactions, and it can
gitudinal Study (NYLS), which was conducted by Thomas, Chess, be differentiated into distress to limitations (anger proneness) and
Birch, Hertzig, and Korn (1963). In their pioneering work, distress to novelty (fearfulness; Sanson, Hemphill, & Smart,
Thomas, Chess, Birch, and colleagues identified nine dimensions 2004). Negative emotionality is considered to be the core dimen-
of temperament: activity level, approach–withdrawal, adaptability, sion of the difficult temperament concept (Bates, 1989; Lee &
mood, threshold, intensity, distractibility, rhythmicity, and atten- Bates, 1985; Prior, 1992; Shiner, 1998). A much-cited definition of
tion span–persistence (Rothbart & Bates, 1998). They also devel- negative emotionality was formulated by Rothbart, Ahadi, and
oped a “difficult” temperament concept in which they included the Hershey (1994), who defined it as the child’s tendency to react to
negative poles of the dimensions approach, adaptability, mood,
stressors with high degrees of emotionality, including anger, irri-
intensity, and rhythmicity. The difficult constellation of tempera-
tability, fear, or sadness.
ment characterized 10% of the NYLS sample (Thomas & Chess,
It frequently has been hypothesized that there is an association
1977).
between temperament and parenting. The intuitive idea is that
Difficult children in the NYLS were found to be at increased
negative emotionality is a temperament characteristic that makes
risk for later behavioral and emotional problems (Thomas, Chess,
children harder to parent (e.g., Chess & Thomas, 1984). The
& Birch, 1968). Although Thomas and Chess’s (1977) conceptu-
attractiveness of this hypothesis is illustrated by the fact that the
alization of difficulty later was criticized in the literature for
focus of studies investigating links between temperament and
methodological as well as conceptual reasons, it has been, and still
parenting has been on attributes falling under the rubric of negative
emotionality (e.g., irritability, difficultness, inhibition, and anger
proneness; Putnam, Sanson, & Rothbart, 2002; Sanson et al.,
Marja C. Paulussen-Hoogeboom, Geert Jan J. M. Stams, Jo M. A. 2004). More adaptable, easy-to-soothe, or sociable children are
Hermanns, and Thea T. D. Peetsma, Department of Education, Faculty of expected to elicit warm and responsive parenting, whereas more
Social and Behavioural Sciences, University of Amsterdam, Amsterdam,
irritable, demanding, or withdrawn children are expected to elicit
The Netherlands.
Correspondence concerning this article should be addressed to Marja C. parental irritation and withdrawal of contact (Putnam et al., 2002).
Paulussen-Hoogeboom, Department of Education, Faculty of Social and Be- The first comprehensive narrative review of studies investigat-
havioural Sciences, University of Amsterdam, P.O. Box 94208, 1090 GE ing the association between negative emotionality and parenting
Amsterdam, The Netherlands. E-mail: m.c.paulussen-hoogeboom@uva.nl was published by Crockenberg (1986). Crockenberg reviewed 16
438
CHILD NEGATIVE EMOTIONALITY AND PARENTING 439

studies testing the hypothesis that negative emotionality in infancy al., 1997; Owens et al., 1998). According to Putnam et al. (2002),
predicts maternal sensitive responsiveness. Nine studies revealed the majority of the studies considering Gender ⫻ Temperament
that babies identified as irritable or as having difficult tempera- interactions found less parental acceptance of irritability and neg-
ments experienced less responsive caregiving or less stimulating ative affect in girls than in boys. There also have been, however,
contact with their mothers. Several studies, however, indicated findings contrary to this pattern (e.g., Crockenberg, 1986; Mac-
mothers of difficult or irritable babies to be more interactive with coby et al., 1984).
their babies. Crockenberg suggested that the conflicting study It is possible that mothers from more resourceful, higher socio-
findings result from interactions between temperament and char- economic backgrounds are more sensitive to the individuality of
acteristics of the caregiver or caregiving environment. In other their children than mothers from less resourceful, lower socioeco-
words, characteristics of the sample under study may determine the nomic backgrounds (M. H. Bornstein, Hahn, Suwalsky, & Haynes,
direction of the association between negative emotionality and 2003). In line with this suggestion, Prior, Sanson, Carroll, and
parenting. For example, when mothers are able (by virtue of their Oberklaid (1989) explored relations between child temperament
personalities, attitudes, and life circumstances) to experience their and parenting in a high- and low-SES group, and they interpreted
baby’s negative emotionality as a challenge, the correlation be- their results as showing that mothers from higher socioeconomic
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

tween negative emotionality and sensitive responsive parenting backgrounds were relatively more sensitive to their children’s
could be positive. On the other hand, when mothers are not so well characteristics. The results, however, appear to have been over-
equipped, the correlation could be negative (Crockenberg, 1986). stated by Prior et al. and should be considered far from clear. In
Findings from studies investigating the relation between child addition, Bates et al. (1982) found no consistent interactions be-
negative emotionality and parenting conducted after Crocken- tween temperament and SES in parent– child relationships.
berg’s (1986) review are comparable with the findings discussed in Finally, the various strategies for measuring negative emotion-
Crockenberg’s original review (Crockenberg & Leerkes, 2003), ality and parenting also may have been responsible for mixed
with most studies reporting negative associations between negative results (Crockenberg & Leerkes, 2003). In a study by Seifer,
emotionality and sensitive responsive parenting (e.g., R. Clark, Schiller, Sameroff, Resnick, and Riordan (1996), for example,
Hyde, Essex, & Klein, 1997; Katainen, Räikkönen, & Keltikangas- relations between infant temperament and maternal responsiveness
Järvinen, 1997; Kiang, Moreno, & Robinson, 2004; Spangler, were only found when negative emotionality was measured by
1990), some studies finding positive associations (e.g., Washing- observation and not when it was measured by maternal report.
ton, Minde, & Goldberg, 1986; Zahr, 1991), and some studies Likewise, Owens et al. (1998) found concurrent negative associ-
finding no association (e.g., Hagekull, Bohlin, & Rydell, 1997; ations between infant irritability and maternal sensitive respon-
Karraker, Lake, & Parry, 1994). Although most researchers con- siveness, which tended to be larger when irritability was measured
tinued to test and report only main effects (Crockenberg & observationally rather than by maternal report.
Leerkes, 2003), these contradictory findings can be considered to It must be noted that most studies examining the association
lend at least some support to Crockenberg’s interaction hypothesis. between negative emotionality and parenting fail to covary mater-
For example, positive associations between negative emotionality nal characteristics, which is an important shortcoming. It renders
and sensitive responsive parenting are most likely to be found in most of the reported associations between negative emotionality
special samples, such as preterm infants (e.g., Washington et al., and parenting inherently questionable, because the possibility ex-
1986; Zahr, 1991). Moreover, in a recent review, Putnam et al. ists that maternal characteristics, such as personality traits, psy-
(2002) suggested that study characteristics (e.g., sample size) and chological functioning, or the amount of support mothers receive
sample characteristics (child’s age, child’s gender, parental char- from the environment, may have influenced their perception of
acteristics, and social factors) may function as possible moderators their children’s negative emotionality (e.g., Rothbart & Bates,
of the link between negative emotionality and parenting. Likewise, 1998) as well as their interactions with the child (e.g., Belsky,
Owens, Shaw, and Vondra (1998) suggested that mixed results are 1984).
due, in part, to methodological differences among studies as well It can be concluded that the majority of studies was particularly
as sample characteristics, including measurement strategies, fam- aimed at investigating the effects of negative emotionality on
ily socioeconomic status (SES), child age and gender, and level of maternal supportive parenting. Support is one of two dimensions
maternal social support. that have been consistently used to describe parenting behavior
How can characteristics of the samples who are studied and (Maccoby & Martin, 1983; Rollins & Thomas, 1979). Support can
characteristics of the studies themselves be expected to influence be described as parental behavior that makes the child feel com-
the association between negative emotionality and parenting? fortable in the relationship with his or her parent and fosters an
Some possibilities are described below. Concerning the child’s internal representation in the child that he or she is basically
age, Crockenberg (1986) suggested that some mothers initially accepted (Rollins & Thomas, 1979). Supportive parenting relates
may respond to their harder-to-parent infants with greater efforts to constructs such as warmth, sensitivity, responsiveness, and
but that they cannot maintain these efforts over time. Study find- acceptance, and it is considered essential for the formation of
ings that are consistent with this notion are those by Peters-Martin secure attachments and other positive developmental outcomes
and Wachs (1984), Maccoby, Snow, and Jacklin (1984), and Bates, (e.g., Coplan, Hastings, Lagace-Seguin, & Moulton, 2002; van
Olson, Pettit, and Bayles (1982), who found positive relations IJzendoorn & Bakermans-Kranenburg, 2004), whereas lack of
between negative emotionality and supportive parenting to be support may contribute to behavior problems (Rothbaum & Weisz,
reversed when the child had grown older. 1994).
The evidence for a role of the child’s gender in the association The second dimension used to describe parenting behavior is the
between negative emotionality and parenting is mixed (Katainen et control dimension. Control attempts can be described as “behavior
440 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

of the parent toward the child with the intent of directing the and quantitative findings of each study are coded; and finally, the
behavior of the child in a manner desirable to the parents” (Rollins resulting data are analyzed using adaptations of conventional sta-
& Thomas, 1979, p. 321). Control strategies may vary from more tistical techniques, such as analysis of variance. Meta-analysis
positive to negative, depending on the parent and the situation. A results in enhanced power, which permits drawing conclusions that
negative strategy that is identified in the literature is restrictive are more accurate than the conclusions presented in any of the
control. This type of control is characterized by high power asser- separate studies (Durlak & Lipsey, 1991). Moreover, meta-
tion, negativity, intrusiveness, hostility, overcontrolling behavior, analysis can be used to discover consistencies in a set of seemingly
or overinvolvement. The extent to which parents use restrictive inconsistent research findings, focusing on the direction and mag-
control has been found to be associated with externalizing behav- nitude of effects across studies. Finally, meta-analysis enables
ior problems (Calkins, 2002; Campbell, 1995). It must be noted researchers to objectively test specific hypotheses.
that restrictive control, especially when measured by observation, The goal of our meta-analytic review was twofold. First, we
does not necessarily involve elements of harsh discipline such as examined the relation between negative emotionality and different
spanking (Bates, Pettit, Dodge, & Ridge, 1998). Another more types of parenting. Second, we examined factors that may function
positive type of control strategy is inductive control. Contrary to as moderators in the linkage between negative emotionality and
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

power assertive methods, inductive control often contains a mes- parenting. We hypothesized that increased negative emotionality
sage about alternative appropriate behavior, and it is characterized would be associated with lower levels of supportive parenting,
by reasoning, reminders of rules, and explanations for the impact higher levels of restrictive control, and lower levels of inductive
of the child’s behavior on others (Kerr, Lopez, Olson, & Sameroff, control. We examined the following sample characteristics and
2004). Inductive control strategies are considered to be effective in study characteristics as possible moderators: age and gender of the
promoting internalization of rules and thereby children’s moral child, birth order, SES, cultural background, maternal age, family
development (Kerr et al., 2004; Maccoby & Martin, 1983). composition, publication year of the studies, impact factor of the
The emphasis on supportive parenting in studies investigating journal in which studies were published, type of study, type of
the association between negative emotionality and parenting is negative emotionality that was measured, and the (combinations
understandable because the majority of research has been directed of) methods used to assess negative emotionality and parenting. To
toward infants. However, when children grow older—and parents address the problem of maternal characteristics as a possible
start to enforce initial standards of behavior and expect children to confounder and source of inflated correlations between negative
comply with their demands—parental control strategies become emotionality and parenting, we performed additional analyses on
increasingly important. studies that provide information about maternal characteristics.
In the same way as negative emotionality is expected to be related
to less supportive parenting, negative emotionality is expected to be
Method
related to higher levels of restrictive control and lower levels of
inductive control. Rubin, Stewart, and Chen (1995) found that early Selection of Studies
social fearfulness and inhibition may elicit restrictive control, that is,
parenting responses that are overprotective and overcontrolling. In This meta-analytic review includes studies published in journal
addition, Katainen et al. (1997) found concurrent as well as longitu- articles until February 2005. The primary search method involved
dinal associations between negative emotionality (i.e., angry aggres- inspection of the computerized databases PsycINFO, ERIC (Edu-
sion and aggressive outbursts) and hostile child-rearing attitudes. cational Resources Information Centre), and PubMed. Combina-
Furthermore, Lee and Bates (1985) observed mothers with 24-month- tions of the following keywords related to negative emotionality
old children, who rated their children as difficult, to use intrusive were used in the searches: temperament, negative emotionality,
control tactics (removing or restraining the child) more frequently difficult temperament, distress, irritability, soothability, discom-
than mothers who rated their children as easy or average. Several fort, anger, fear, frustration, and sadness. With regard to parent-
studies found relations between negative emotionality and inductive ing, the following keywords were used: parenting, parental,
control to be in the expected direction. Harrington, Black, Starr, and mother*, father*, maternal, paternal, parent– child, childrearing,
Dubowitz (1998) found that difficultness in 1-year-olds was related to and care giving. An asterisk indicates that the search was not
lower levels of positive reinforcement and praise from the mother. limited to the particular word or fragment. Additional studies were
Calkins (2002) found 1-year-olds’ distress reactions to laboratory located using the ancestry method (inspection of the references
frustration tasks to be negatively associated with maternal positive section of obtained studies and relevant narrative reviews).
guidance. Studies had to meet several criteria to be included in the meta-
To summarize, the results of studies investigating the link be- analyses. First, the study had to be published in a peer-reviewed
tween negative emotionality and parenting call for a convincing journal. This approach provides some degree of quality control but
answer to the question of whether negative emotionality is related risks publication bias because journals may be selective in pub-
to parenting behavior. If so, a second question would be whether lishing only results that are characterized by lower p values and
there are specific conditions or patterns of conditions under which larger effect sizes (Rosenthal, 1995). We followed Rosenthal’s
the direction or the strength of the association is altered. Both (1995) recommendations for detecting possible publication bias
questions can be answered through meta-analysis. Meta-analysis is (see below). Second, the study had to use measures of negative
a statistical technique enabling researchers to quantitatively sum- emotionality and parenting. Third, only studies covering infancy
marize the findings of separate empirical studies according to a and preschool age were included. Fourth, the study had to be
three-step procedure (Lipsey & Wilson, 2000). First, a sample of conducted in a Western culture (United States, Europe, or Austra-
eligible studies is gathered; second, the appropriate characteristics lia). Non-Western societies were excluded because of assumed
CHILD NEGATIVE EMOTIONALITY AND PARENTING 441

differences between Western and non-Western cultures in both measured children’s angry distress (e.g., Calkins, Hungerford, &
negative emotionality and parenting (R. M. Lerner, Rothbaum, Dedmon, 2004), type of negative emotionality was coded as anger.
Boulos, & Castellino, 2002). Fifth, studies of clinical or other When a study measured children’s fearful distress (including in-
special samples of children (e.g., adopted children, children born hibition; e.g., Kochanska, 1998), type of negative emotionality was
prematurely) or parents (e.g., parents with alcohol or drug prob- coded as fear. When a study measured both angry and fearful
lems, adolescent mothers) were not included. Extreme groups may distress (e.g., Belsky, Hsieh, & Crnic, 1998), type of negative
yield effects of larger magnitude or different direction from what emotionality was coded as anger and fear mixed. When studies
would be expected in a general population sample (Rothbaum & used a difficult temperament construct similar to that of Thomas
Weisz, 1994). Finally, samples included in the meta-analysis had and Chess (1977; e.g., Maziade et al., 1989), type of negative
to be statistically independent. When studies were identified to be emotionality was coded as difficultness. When studies measured
statistically dependent, the oldest, original study was included in Thomas and Chess’s mood dimension (e.g., Mangelsdorf, Gunnar,
the analysis. Exceptions were made when the later study provided Kestenbaum, Lang, & Andreas, 1990), type of negative emotion-
more extensive information on the same measurement moment or ality was coded as negative mood. Finally, some studies reported
when the aims of the later study were more closely related to the on negative reactivity or irritability, especially in neonates (e.g.,
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

aim of this meta-analysis (i.e., examination of the association Worobey, Laub, & Schilmoeller, 1983). In these cases, type of
between negative emotionality and parenting). negative emotionality was coded as irritability.
The large majority of studies investigating associations between SES was coded as low, middle, or high. As a rule, we followed
temperament and parenting are cross-sectional. In case of longi- the SES classification reported in the studies. If no SES classifi-
tudinal designs, the decision was made to include only statistics cation was available, the following criteria were applied. Samples
from the first time of measurement. In this way, study statistics were coded as low SES in studies in which (a) a substantial
were least likely to be influenced by previous measurements. number of participants was reported to be poor; (b) participants
Sometimes, however, it was impossible to extract concurrent sta- were recruited from hospitals predominantly serving low-income
tistics from longitudinal studies. In those cases, we chose to use the families; (c) a substantial number of the participants was reported
closest possible time range that was provided by the study. In 9 of to be enrolled in support programs, prevention programs, or food
the 10 longitudinal studies, this time range was between 3 and 13 programs; or (d) a substantial number of the sample was reported
months. A rather long time range (50.5 months) was encountered as not having a middle or high school degree (which usually meant
in the study by Maziade, Cote, Bernier, Boutin, and Thivierge an average level of maternal education of 12 years or less).
(1989). When studies did not provide enough statistical informa- Samples were coded as middle SES when (a) the participants were
tion for the calculation of effect sizes, authors were contacted. reported to include a broad range of SES in terms of education,
The literature search resulted in 62 studies that met the inclusion occupation, and income; (b) the sample was reported to not be
criteria. A detailed list of the included studies is provided in Table markedly deviant from expected distributions on key demographic
1. Initially, 94 studies addressing the association between negative variables, such as education, occupation, and income; (c) average
emotionality and parenting were obtained. However, 32 studies level of maternal education was between 12 and 15 years; or (d)
were excluded given that samples were statistically dependent or the average family income was indicative of middle SES. Samples
because it was not possible to derive statistical data relevant for the were coded as high SES when (a) the sample was described as
calculation of effect sizes (the authors could not be contacted or highly educated (average level of education was 15 years or more),
they did not have access anymore to their original data). (b) occupational ratings were skewed toward the upper SES levels,
and (c) the average family income reflected higher SES. In some
Coding the Studies cases, studies reported participants’ scores on Hollingshead’s orig-
inal Two Factor Index of Social Status (Hollingshead & Redlich,
A coding system was used to rate every study on design and 1958) or the later Hollingshead Four Factor Index of Social Status
measurement characteristics as well as on sample and publication (M. H. Bornstein et al., 2003; Hollingshead, 1975). Concerning
characteristics. The following design and measurement character- these indices, it should be noted that Hollingshead’s original two
istics were included: sample size, type of study (cross-sectional or factor index ranges from 11 to 77 and has a reversed scale, with
longitudinal), type of negative emotionality (see below), methods higher scores indicating lower SES. Hollingshead’s four factor
used to assess negative emotionality and parenting (parent report, index, on the other hand, ranges from 8 to 66, with higher scores
observation, or composite), and combination of assessment meth- indicating higher SES.
ods used for negative emotionality and parenting, respectively Reported parenting behaviors were coded into three types. Par-
(parent report–parent report, observation– observation, parent enting behavior was categorized as supportive when it involved
report– observation, and composite). Age of the child, gender of constructs like responsiveness, sensitivity, involvement, syn-
the child (percentage of boys), birth order (percentage of first-born chrony between parent and child, approval, or affection (e.g.,
children), age of mothers, SES (see below), ethnicity (percentage Keefe, Kotzer, Froese-Fretz, & Curtin, 1996; Kochanska, 1998).
of Caucasian White mother– child dyads in the sample), and family Parenting behavior was classified as restrictive when it was char-
composition (percentage of two-parent families) were included as acterized by high power assertion, negativity, intrusiveness, hos-
sample characteristics. For publication characteristics, we coded tility, overcontrolling behavior, or overinvolvement (e.g., Lehman,
year of publication and impact factor of the journal (Thomson Steier, Guidash, & Wanna, 2002; Rubin, Burgess, & Hastings,
Corporation, 2003). 2002). Parenting was classified as inductive when it included
As expected, the concept of negative emotionality was opera- authoritative limit setting and appeals to the use of reason, such as
tionally defined in diverse ways across studies. When a study offering reminders of rules and explanations for the impact of the
442 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

Table 1
Sample Characteristics, Type of Negative Emotionality, and Parenting Category of the Studies Included in the Meta-Analyses

Sample characteristics

% Agea % % Type of
Age child % first mother Caucasian intact negative
Study N (months) male born (years) SES White families emotionality Parenting category

Bates, Olson, Pettit, & Bayles (1982) 168 6 56 51 M Difficultness Support


Belsky, Hsieh, & Crnic (1998) 125 11.5–15 100 100 29 M 100 100 Anger and fear Support
mixed Restrictive control
Calkins, Hungerford, & Dedmon 162 6 49 29 M 80 80 Anger Support
(2004) Restrictive control
Calkins & Johnson (1998) 73 18 56 M 89 Anger Support
Restrictive control
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Campbell (1979) 12 3 33 50 26.5 M 100 100 Difficultness Support


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Clark, Kochanska, & Ready (2000) 112 9 50 31 M 97 100 Anger and fear Support
mixed Restrictive control
Clark, Hyde, Essex, & Klein (1997) 198 4 46 50.5 30 M 92.4 100 Irritability Support
Restrictive control
Clarke-Stewart (1973) 36 13.5 50 100 22 L 50 61 Irritability Support
Restrictive control
Coffman, Levitt, Guacci, & Silver 51 13 59 44 31 M 98 90 Difficultness Support
(1992) Restrictive control
Coplan, Prakash, O’Neil, & Armer 246 50 50 M 82 Fear Support
(2004) Restrictive control
Inductive control
Crockenberg & Smith (1982) 56 0–3 50 66 26 M 96 100 Irritability Support
Daniels, Plomin, & Greenhalgh 120 12 M 100 Difficultness Support
(1984) Restrictive control
Dunn & Kendrick (1980) 40 25 53 100 M Negative mood Restrictive control
Eisenberg & Fabes (1994) 79 61.5 53 H 91 100 Fear Support
Restrictive control
Inductive control
Fish & Stifter (1993) 87 5 52 49 29 H 98 87 Anger Support
Restrictive control
Fish, Stifter, & Belsky (1991) 83 0–5 52 29 M 98 87 Irritability Support
Restrictive control
Gauvain & Fagot (1995) 26 22–33 46 M 100 Difficultness Support
Gauvain & Fagot (1995) 26 18–31.2 54 M 100 Difficultness Support
Gordon (1983) 74 42 47 57 M 75–100 Difficultness Restrictive control
Inductive control
Hagekull & Bohlin (1986) 30 14.3 53 67 29.5 H 100 97 Difficultness Support
Restrictive control
Hagekull, Bohlin, & Rydell (1997) 115 7 51 39 H 100 98 Difficultness Support
Halpern & McLean (1997) 20 4 50 35 23 L 55 60 Difficultness Support
Hann (1989) 34 5 67 47 27.5 M Difficultness Support
Harrington, Black, Starr, & Dubowitz 121 18.3 36 24.5 L 5 12 Difficultness Support
(1998)
Houldin (1987) 20 5.2 55 60 M 90 Negative mood Support
Karraker, Lake, & Parry (1994) 60 10.5 50 45 30.5 M 100 Difficultness Support
Inductive control
Katainen, Räikkönen, & Keltikangas- 577 36 51 M 100 Anger Restrictive control
Järvinen (1997)
Keefe, Kotzer, Froese-Fretz, & 40 1.5 50 60 30 M 75 100 Irritability Support
Curtin (1996)
Kiang, Moreno, & Robinson (2004) 175 6–14.5 50 100 20 L 31 Difficultness Support
Kochanska (1995) 103 33 50 M 80 Fear Inductive control
Kochanska (1998) 112 9 50 39 M 97 100 Fear Support
Kochanska, Friesenborg, Lange, & 102 7 50 42 31 M 91 100 Anger and fear Support
Martel (2004) mixed
Kyrios & Prior (1990) 120 44 49 47 32.5 H 96 Difficultness Restrictive control
Leerkes & Crockenberg (2003) 90 6 56 100 29 M 93 99 Anger and fear Support
mixed
Lehman, Steier, Guidash, & Wanna 51 21 50 33 H 100 75–100 Anger and fear Support
(2002) mixed Restrictive control
Lerner & Galambos (1985) 89 24–36 54 M 92 Difficultness Support
Luster, Boger, & Hannan (1993) 544 6 25 L 56 77 Irritability Support
CHILD NEGATIVE EMOTIONALITY AND PARENTING 443

Table 1 (continued )

Sample characteristics

% Agea % % Type of
Age child % first mother Caucasian intact negative
Study N (months) male born (years) SES White families emotionality Parenting category

Maccoby, Snow, & Jacklin (1984) 57 12 49 M Difficultness Support


Restrictive control
Mangelsdorf, Gunnar, Kestenbaum, 66 9 47 56 M 100 Negative mood Support
Lang, & Andreas (1990)
Mangelsdorf, McHale, Diener, Heim 102 8 54 37 32 H Anger and fear Support
Goldstein, & Lehn (2000) mixed
Maziade, Cote, Bernier, Boutin, & 80 6 62 M 75–100 Difficultness Inductive control
Thivierge (1989)
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Mertesacker, Bade, Haverkock, & 37 4 51 100 28 M 100 Negative mood Support


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Pauli-Pott (2004)
Milliones (1978) 24 11.5 46 24 L 0 Difficultness Support
Moss (1967) 30 1 50 100 Irritability Support
Nachmias, Gunnar, Mangelsdorf, 78 18 51 50 32.5 H 100 Fear Support
Parritz, & Buss (1996)
NICHD Early Child Care Research 1364 6 52 M 76 86 Difficultness Support
Networkb (1999)
Owens, Shaw, & Vondra (1998) 235 12 55 26 L 58 57 Irritability Support
Pauli-Pott, Mertesacker, Bade, Bauer, 101 4 57 100 29 M 100 Anger and fear Support
& Beckmann (2000) mixed
Peters-Martin & Wachs (1984) 30 12 53 100 H 97 100 Difficultness Support
Prior, Sanson, Carroll, & Oberklaid 47 40 50 H Negative mood Support
(1989)
Rothbart (1986) 46 3 50 M 75–100 Anger and fear Support
mixed
Rubin, Burgess, & Hastings (2002) 108 25 50 31 M 97 100 Fear Restrictive control
Schuler, Black, & Starr (1995) 20 1 60 24.5 L 10 10 Irritability Support
Seifer, Schiller, Sameroff, Resnick, 49 6 49 29 H 100 100 Anger and fear Support
& Riordan (1996) mixed
Shaw & Vondra (1995) 100 12–15 59 25 L 61 46 Difficultness Support
Simonds & Simonds (1981) 182 55 60 53 M 100 93 Difficultness Support
Restrictive control
Inductive control
Simpson & Stevenson-Hinde (1985) 47 42 55 M Anger and fear Support
mixed
Spangler (1990) 24 12–15.5 54 38 M 100 Difficultness Support
Susman-Stillmann, Kalkoske, 267 3 100 20.5 L 80 62 Irritability Support
Egeland, & Waldman (1996)
van der Mark, Bakermans- 131 16 0 100 32.5 H 95 Fear Support
Kranenburg, & van IJzendoorn Restrictive control
(2002) Inductive control
Vaughn, Taraldson, Crichton, & 187 6 55 100 20.5 L 80.5 38 Difficultness Support
Egeland (1981)
Worobey, Laub, & Schilmoeller 24 1 50 38 27.5 M 100 100 Irritability Support
(1983)

Note. SES ⫽ socioeconomic status; L ⫽ low; M ⫽ middle; H ⫽ high.


a
Age of mother is rounded off to 0.5 years. b The NICHD Early Child Care Research Network (1999) study was winsorized (n ⫽ 544).

child’s behavior on others (e.g., Simonds & Simonds, 1981; van sents the magnitude of the association between two variables, regard-
der Mark, Bakermans-Kranenburg, & van IJzendoorn, 2002). less of sample size. In our study, relevant outcomes were effect sizes
The studies were coded by Marja C. Paulussen-Hoogeboom and for the association between negative emotionality and parenting be-
Geert Jan J. M. Stams. The intercoder reliability of the coding haviors (support, restrictive control, and inductive control). When a
system was established on 15 studies. In all cases, reliabilities were study did not report the correlation coefficient between negative
satisfactory (␬ ⬎ .90). emotionality and parenting itself but only the nonsignificance of the
association, an effect size of zero was assigned. This is a commonly
used but conservative strategy, which generally underestimates the
Meta-Analytical Procedures and Decisions
true magnitude of effect sizes (Durlak & Lipsey, 1991). Assigning an
In a meta-analysis, study outcomes are the unit of analysis. In the effect size of zero is preferred, however, to exclusion of the nonsig-
present meta-analyses, Pearson’s product–moment correlation coeffi- nificant results from the meta-analyses, as this would result in an
cient (r) was used as the effect size estimate. This effect size repre- overestimation of combined effect sizes (Rosenthal, 1995).
444 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

A number of studies (e.g., Spangler, 1990; Vaughn, Taraldson, reporting on these meta-analyses, we decided to present the com-
Crichton, & Egeland, 1981) examined different variables within bined effect sizes and their CIs in terms of a random-effects model.
the same parenting category. When this was the case, correlations In the case of heterogeneity, random-effects model parameters are
were averaged resulting in one effect size for the particular par- more conservative than fixed-effects model parameters, and the
enting category. In one case (Gauvain & Fagot, 1995), two sam- moderator tests (which are based on the fixed effects) should be
ples were described in one study. These samples were considered considered to be descriptive only of the specific set of studies at
as separate studies. The relatively large sample size of the National hand and thus should be interpreted with caution (Rosenthal, 1995;
Institute of Child Health and Human Development (NICHD) Early van IJzendoorn, Vereijken, Bakermans-Kranenburg, & Riksen-
Child Care Research Network (1999) study (N ⫽ 1,364) was set Walraven, 2004).
equal to the next largest sample size (N ⫽ 544; Luster, Boger, & Moderator tests were performed to explain the variability be-
Hannan, 1993). This procedure is known as winsorizing, and it is tween effect sizes, using categorical testing procedures as de-
performed in order to prevent the meta-analytic results from being scribed by Lipsey and Wilson (2000). These procedures are anal-
overly determined by one disproportional large sample size (Ham- ogous to analysis of variance. Categorical testing yields two
pel, Ronchetti, Rousseeuw, & Stahel, 1986). Outlying effect sizes
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

homogeneity estimates, a within-groups Q (Qw) and a between-


This document is copyrighted by the American Psychological Association or one of its allied publishers.

were identified on the basis of z values larger than 3.3 or smaller groups Q (Qb). A significant Qw indicates that the effect sizes
than ⫺3.3 (Tabachnick & Fidell, 2001). within each moderator variable category are heterogeneous. A
significant Qb indicates that the subgroups of effect sizes are
Data Analysis significantly different from one another.
However, simple grouping variables may not be sufficient to
The resulting three sets of independent effect sizes (no study or account fully for between-studies heterogeneity in many meta-
participant was counted more than once in each meta-analysis) analyses (Durlak & Lipsey, 1991). Therefore, when allowed by the
were inserted in M. Borenstein, Rothstein, and Cohen’s (2000) ratio of predictors to studies, we also used a multiple regression
comprehensive meta-analysis (CMA) program. Before computing format. In the regression analysis, moderators that were identified
combined effect sizes, we used the CMA to weight individual as significant in the categorical testing procedure were entered as
effect sizes by sample size, resulting in unbiased effect sizes. This
predictor variables to test whether they accounted for unique
is done because in meta-analysis the effect sizes from various
variance in the unbiased effect size r.
studies are based on different sample sizes. Consequently, the
homogeneity of variance, required for conventional statistics, is
lacking. Notably, correlations become more stable as sample size Publication Bias (File Drawer Analysis)
increases, and effect sizes based on large samples deviate less from
the population effect size than those based on smaller samples A common problem in conducting meta-analysis is that many
(Mullen, 1989). CMA computed fixed- as well as random-effects studies remain unpublished because of nonsignificant findings. To
model parameters and 95% confidence intervals (CIs) around the examine whether such publication bias or file drawer problem
point estimate of the combined effect sizes. Significance tests and exists, one can calculate the fail-safe number, which indicates the
moderator analyses in fixed-effects models are based on the as- minimum number of additional studies with nonsignificant results
sumption that differences among studies leading to differences in that is needed to reduce significant meta-analytic results to non-
effect sizes are not random and that, in principle, the set of study significance (Durlak & Lipsey, 1991). Meta-analytic findings are
effect sizes is homogeneous at the population level (Lipsey & considered to be robust if the fail-safe number exceeds the critical
Wilson, 2000). Significance testing in fixed-effects models is value obtained with Rosenthal’s (1995) formula of (5 ⫻ k) ⫹ 10
based on the total number of participants in the meta-analysis, but in which k is the number of studies used in the meta-analysis. If the
generalization is limited to other participants who might have been fail-safe number falls below this critical value, a publication bias
included in the same studies of the meta-analysis (Rosenthal, or file drawer problem may exist. Appropriate fail-safe numbers
1995). In random-effects models, homogeneity at the population ( p ⬍ .05) for our meta-analyses were computed using SPSS
level is not assumed (Lipsey & Wilson, 2000). Significance testing macros from Lipsey and Wilson (2000). Notably, larger mean
in random-effects models is only based on the total number of effect sizes yield increased fail-safe numbers, as the fail-safe
studies in the meta-analysis, and generalization is limited to the number is based on a fraction formula, with the mean effect size in
population of studies from which the current set of studies was the numerator and the criterion effect size level in the denominator.
drawn (Rosenthal, 1995).
Random effects models are presumed to mirror the heterogene-
ity that is found in behavioral studies more adequately than fixed- Results
effects models and use noninflated alpha levels when the require-
ment of homogeneity has not been met (Hedges & Olkin, 1985). The 62 studies included in this meta-analytic review reported
Homogeneity analyses (Q statistics) were conducted to test the data on 7,613 mother– child dyads. The distributions of effect sizes
assumption that the sets of effect sizes in the three separate for the three categories of parenting behavior are presented in
meta-analyses were homogeneous at the population level and to Table 2. No outlying effect sizes were identified. In interpreting
test the significance of moderators (M. Borenstein et al., 2000; the magnitude of the combined effect sizes, we used generally
Mullen, 1989; Rosenthal, 1995). In two cases (the meta-analysis accepted conventions formulated by Cohen (1988). Effect sizes of
on supportive parenting and the meta-analysis on restrictive con- r ⱕ .10, r ⫽ .25, and r ⱖ .40 were considered as indices of small,
trol), the set of effect sizes was found to be heterogeneous. In medium, and large effects, respectively (Lipsey & Wilson, 2000).
CHILD NEGATIVE EMOTIONALITY AND PARENTING 445

Table 2
Stem and Leaf Display of Effect Sizes for the Associations Among Support, Inductive Control,
Restrictive Control, and Negative Emotionality (r)

Leaf

Stem Support Restrictive control Inductive control

.7 0
.6
.5 4 0
.4 2, 3, 4, 4
.3 8 1, 3, 3 3
.2 2, 2, 4, 5, 5, 7, 7 2, 3, 3
.1 2, 5, 5, 5, 5, 5, 7, 9 0, 4 7, 9, 9
.0 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 4, 5, 5, 6, 6, 6, 6, 8 0, 0, 0, 0, 2, 3, 5, 6, 6 0, 0, 0, 0, 3
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

⫺.1 0, 2, 9 0, 7, 7
This document is copyrighted by the American Psychological Association or one of its allied publishers.

⫺.2 2 7
⫺.3 0, 1, 4, 5
⫺.4 4, 6
⫺.5 2

Note. The stem contains the first digit of the correlation coefficient, whereas the leaf contains the second digit.
In the comprehensive meta-analysis program, effect sizes that are in the expected direction have a positive sign.

Supportive Parenting The combined effect size for the 11 (n ⫽ 1,720) studies using
low-SES samples (r ⫽ .19, p ⬍ .001) was larger than the combined
In the meta-analysis of the association between negative emo-
effect size for the 32 studies (n ⫽ 2,954) using middle-class
tionality and supportive parenting, 55 studies were included, in-
samples (r ⫽ .08, p ⬍ .01). However, the combined effect size for
volving 5,467 mother– child dyads (after winsorizing). We found a
the 11 studies (n ⫽ 764) using high-SES samples (r ⫽ ⫺.11, p ⬍
small combined effect size of r ⫽ .061 (CI ⫽ .01–.11, random
.05) was in the unexpected direction, meaning that in higher SES
model, p ⬍ .05), indicating higher levels of negative emotionality
samples, higher levels of negative emotionality in children are
to be weakly but significantly related to less supportive parenting.
associated with more instead of less supportive parenting. Thus,
However, a file drawer problem was encountered. The random
model fail-safe number for supportive parenting (k ⫽ 55) was 45, increased negative emotionality was relatively strongly associated
which is below Rosenthal’s (1995) critical value of (55 ⫻ 5) ⫹ with less supportive parenting in low-SES samples, whereas a
10 ⫽ 285. This indicates that publication bias could exist, because smaller association was found in middle-SES samples, and an even
only 45 unpublished studies reporting no associations between reversed association was found in high-SES samples.
negative emotionality and supportive parenting were needed to The association between more negative emotionality and less sup-
reduce this combined effect size to nonsignificance. The set of portive parenting was only significant for studies using mixed sam-
unbiased effect sizes was heterogeneous, Qw(54) ⫽ 167.57, p ⬍ ples (less than 75% Caucasian White participants). The combined
.001, indicating that differences in effect sizes among studies may effect size for the 9 studies (n ⫽ 1,275) using mixed samples was r ⫽
be explained by moderators, such as sample or design character- .21, p ⬍ .001, whereas the combined effect size for the 33 studies
istics. (n ⫽ 3,395) using predominantly White samples was r ⫽ .02.
Moderator analyses. As the set of effect sizes proved to be The association between more negative emotionality and less
heterogeneous, it was important to conduct moderator analyses in supportive parenting was significant for studies measuring anger
order to explain the variation in effect sizes among studies (r ⫽ .16, p ⬍ .05, k ⫽ 3, n ⫽ 299) and difficultness (r ⫽ .06, p ⬍
(Mullen, 1989). The results of the moderator analyses are pre- .05, k ⫽ 22, n ⫽ 2,148). The combined effect sizes for studies
sented in Table 3. In order to correct for multiple comparisons, measuring fear (r ⫽ ⫺.07, k ⫽ 5, n ⫽ 516), anger and fear mixed
thereby preventing Type I errors caused by inflated alphas, we (r ⫽ .01, k ⫽ 10, n ⫽ 820), irritability (r ⫽ .08, k ⫽ 11, n ⫽
conducted moderator analyses at a significance level of ␣ ⫽ .001. 1,520), and negative mood (r ⫽ .29, k ⫽ 4, n ⫽ 164) were
As can be derived from Table 3, five moderators were significant: nonsignificant.
age of mother, Qb(2) ⫽ 17.60, p ⬍ .001; SES, Qb(2) ⫽ 39.48, p ⬍ The association between negative emotionality and supportive
.001; ethnicity, Qb(1) ⫽ 17.47, p ⬍ .001; type of negative emo- parenting was only significant for studies using parent report to
tionality, Qb(5) ⫽ 21.25, p ⬍ .001; and assessment method of assess negative emotionality (r ⫽ .11, p ⬍ .01, k ⫽ 31, n ⫽ 3,068).
negative emotionality, Qb(2) ⫽ 16.13, p ⬍ .001. The combined effect size for the 11 studies (n ⫽ 763) using
More negative emotionality in children was clearly associated behavioral observation (r ⫽ ⫺.09) and the combined effect size for
with less supportive parenting for mothers younger than 25 years the 13 studies (n ⫽ 1,636) using a composite measure (r ⫽ ⫺.06)
of age (r ⫽ .22, p ⬍ .001, k ⫽ 8, n ⫽ 841). The combined effect was smaller and nonsignificant.
size for the 17 studies (n ⫽ 1,768) with mothers between 25 and
30 years of age and the combined effect size for the 9 studies (n ⫽
869) with mothers older than 30 years of age was smaller and 1
In the CMA program, effect sizes that are in the expected direction
nonsignificant (r ⫽ .07 and r ⫽ ⫺.03, respectively). have a positive sign.
446 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

Table 3
Moderators of Effect Sizes for Studies on Negative Emotionality and Supportive Parenting

Moderator n k r 95% CI Qb Qw
*
Total set 5,467 55 .06 .01 to .11 167.57***
Publication yeara 9.59**
Before 1987 1,067 15 ⫺.04 ⫺.14 to .07 33.79**
From 1987 on 4,400 40 .09** .03 to .15 124.19***
Impact factor 7.14**
⬍2 3,317 35 .07* .00 to .14 111.74***
ⱖ2 2,150 20 .03 ⫺.05 to .11 48.69***
Type of study 1.84
Cross-sectional 4,776 46 .06* .01 to .12 149.43***
Longitudinal 691 9 .02 ⫺.11 to .14 16.31*
Age of childb 8.71*
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Neonates 169 5 ⫺.13 ⫺.52 to .31 25.95***


This document is copyrighted by the American Psychological Association or one of its allied publishers.

Early infancy 3,446 27 .10** .04 to .16 61.79***


1-year-olds 1,179 14 .04 ⫺.09 to .16 54.10***
Preschoolers 673 9 .01 ⫺.12 to .13 17.03*
Gender 6.07*
⬍50% boys 846 10 .11 ⫺.06 to .27 41.41***
ⱖ50% boys 3,710 42 .03 ⫺.03 to .09 107.59***
Percentage first born 3.99*
⬍75% 1,529 21 .03 ⫺.07 to .13 68.85***
75%–100% 1,193 11 .09 ⫺.02 to .19 29.02**
Age of mother 17.60***
⬍25 years 841 8 .22*** .09 to .34 17.47*
25–30 years 1,768 17 .07 ⫺.03 to .17 54.39***
ⱖ30 years 869 9 ⫺.03 ⫺.15 to .09 20.63**
SES 39.48***
Low 1,720 11 .19*** .11 to .27 22.42*
Middle 2,954 32 .08** .01 to .14 70.42***
High 764 11 ⫺.11* ⫺.23 to .01 23.87**
Ethnicity 17.47***
⬍75% White 1,275 9 .21*** .11 to .29 15.08
75%–100% White 3,395 33 .02 ⫺.04 to .08 83.04***
Percentage intact families 8.23**
⬍75% 977 8 .18** .07 to .29 17.83*
75%–100% 3,168 29 .08** .02 to .14 73.62***
Type of NE 21.25***
Anger 299 3 .16* .00 to .30 2.91
Fear 516 5 ⫺.07 .03 to .16 19.23***
Anger and fear mixed 820 10 .01 ⫺.07 to .10 12.24
Irritability 1,520 11 .08 ⫺.06 to .20 45.04***
Difficultness 2,148 22 .06* ⫺.01 to .13 43.57**
Negative mood 164 4 .29 ⫺.22 to .67 23.34***
Assessment method NE 16.13***
Parent report 3,068 31 .11** .04 to .17 74.63***
Observation 763 11 ⫺.09 ⫺.24 to .06 38.17***
Composite 1,636 13 .06 ⫺.04 to .16 38.64***
Assessment method parenting 0.36
Parent report 505 5 .05 ⫺.12 to .21 11.18
Observation 3,670 42 .05 ⫺.12 to .21 139.35***
Composite 1,292 8 .08* ⫺.02 to .18 16.68*
Combination of assessment methods 15.37**
Parent report–parent report 505 5 .05 ⫺.12 to .21 11.18*
Observation–observation 632 10 ⫺.11 ⫺.28 to .07 37.71***
Parent report–observation 2,019 25 .12** .04 to .20 62.13***
Composite 2,311 15 .07* ⫺.01 to .15 41.17***

Note. Unbiased effect sizes (rs) were calculated with one-tailed alpha set at .05. k ⫽ number of studies; CI ⫽
confidence interval; Qb ⫽ homogeneity statistic (between classes); Qw ⫽ homogeneity statistic (within classes);
SES ⫽ socioeconomic status; NE ⫽ negative emotionality.
a
Before 1987 ⫽ before Crockenberg’s (1986) review article; From 1987 on ⫽ after Crockenberg’s (1986)
original review article. b Neonates ⫽ 0 –2 months; early infancy ⫽ 3–11 months; 1-year-olds ⫽ 12–23 months;
preschoolers ⫽ 24 –59 months.
*
p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.
CHILD NEGATIVE EMOTIONALITY AND PARENTING 447

Regression analysis. To test whether the identified moderators unbiased effect sizes was heterogeneous, Qw(21) ⫽ 69.10, p ⬍ .001,
accounting for differences in effect sizes were confounded, we pointing toward the possibility of moderation by other variables. The
performed a multiple regression analysis with unbiased rs as study calculated fail-safe number (k ⫽ 22) was 29 (random model), which
outcomes. Only moderators that were significant at the p ⬍ .001 is below Rosenthal’s (1995) critical value of (22 ⫻ 5) ⫹ 10 ⫽ 120,
level (age of mother, SES, ethnicity, type of negative emotionality, indicating the possibility of publication bias.
and assessment method of negative emotionality) were entered as Moderator analyses. The results of the moderator analyses for
predictors. Missing values were imputed using the expectation this meta-analysis are presented in Table 4. Given that there was
maximalization algorithm (Bernaards & Sijtsma, 1999). Expecta- only one study in this meta-analysis with low-SES participants, we
tion maximalization is considered to be adequate when values are decided to collapse the categories of low and middle SES to
missing at random. Age of mother and ethnicity were already increase statistical power in the moderator analysis. In a similar
coded as continuous variables. SES, type of negative emotionality, way and for the same reason, we decided to combine mothers who
and assessment method of negative emotionality were discrete were less than 25 years old with mothers between 25 and 30 years
variables made suitable for regression analyses by means of of age into one category. Collapsing categories was no solution for
dummy coding (Tabachnick & Fidell, 2001). SES was transformed
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

ethnicity and family composition (discrete variables), as this would


This document is copyrighted by the American Psychological Association or one of its allied publishers.

into a set of two dummy variables: low versus middle– high SES result in variables consisting of only one category, rendering these
and low–middle versus high SES. Type of negative emotionality moderators untestable. Consequently, we decided not to analyze
was recoded into a set of five dummy variables: anger, fear, ethnicity and family composition for lack of statistical power and
irritability, difficultness, and negative mood. Finally, assessment unreliability of results when moderator analyses were based on
method of negative emotionality was transformed into a set of two testing differences between categories containing no more than
dummy variables: composite– observation versus parent report and one study.
composite–parent report versus observation. As can be derived from Table 4, three moderators were signif-
Preliminary analyses showed one of the two dummy variables icant at the p ⬍ .001 level: age of child, Qb(2) ⫽ 26.32, p ⬍ .001;
for SES to be highly correlated with age of mother (r ⫽ .79) and birth order, Qb(1) ⫽ 13.28, p ⬍ .001; and combination of assess-
ethnicity (r ⫽ .82). Age of mother and ethnicity were also highly ment methods, Qb(3) ⫽ 21.16, p ⬍ .001. Higher levels of negative
correlated (r ⫽ .64). One of the dummy variables for assessment emotionality were associated with more restrictive parenting in
method of negative emotionality correlated highly with the dummy infants and in preschoolers but not in 1-year-olds. Combined effect
for difficultness (r ⫽ .66). Apart from this finding, most correla- sizes were larger for the nine studies with preschoolers (n ⫽ 1,353,
tions between moderators were small, with few exceptions of r ⫽ .13, p ⬍ .01) and the five studies with infants (n ⫽ 603, r ⫽
variables that were moderately correlated (range: .01 ⬍ r ⬍ .59). .22, p ⬍ .001) than for the eight studies with 1-year-olds (n ⫽ 603,
All 11 predictors (age of mother, ethnicity, low vs. middle– high r ⫽ ⫺.03). Moderator analysis of birth order showed that the
SES, low–middle vs. high SES, fear, anger, irritability, difficultness, association between negative emotionality and restrictive parent-
negative mood, composite– observation vs. parent report, and ing was only significant in studies with less than 75% first-born
composite–parent report vs. observation) were entered into the regres- children (r ⫽ .19, p ⬍ .01, k ⫽ 7, n ⫽ 719). The combined effect
sion analysis, yielding a significant regression equation, F(11, 54) ⫽ size for the four studies (n ⫽ 332) with more than 75% first-born
52.36, p ⬍ .001. The moderators accounted for 45% of the variance children was nonsignificant (r ⫽ ⫺.01). Finally, the five studies
in effect sizes. Three variables emerged as significant predictors from (n ⫽ 1,080) assessing both negative emotionality and parenting by
the regression analysis, qualifying the relation between negative emo- means of parent self-report as well as the five studies (n ⫽ 695)
tionality and supportive parenting: the dummy variable for low– assessing negative emotionality by means of parent report and
middle versus high SES (b ⫽ ⫺.36, p ⬍ .001), ethnicity (b ⫽ ⫺.35, parenting by means of composite measures yielded significant
p ⬍ .05), and the dummy variable for composite–parent report versus combined effect sizes (r ⫽ .21, p ⬍ .001, and r ⫽ .15, p ⬍ .05,
observation (b ⫽ ⫺.26, p ⬍ .05). Studies using higher SES, predom- respectively). Other combinations of assessment methods yielded
inantly Caucasian White samples, and observation to assess negative much smaller and nonsignificant combined effect sizes.
emotionality had smaller effect sizes than samples of low to middle Regression analysis. The significant moderators, with the ex-
SES, with relatively high percentages of participants from ethnic ception of birth order (50% missing values), were dummy coded
minorities, and than studies using parent report or composite measures and entered into a multiple regression analysis. Both age of child
to assess negative emotionality. We controlled for multicollinearity by and combination of assessment methods emerged from the analy-
adding error variance to each of the nine predictors and subsequently sis as unique moderators of the relation between child negative
running the regression analysis for a second time. The second regres- emotionality and restrictive parenting, accounting for 63% of the
sion analysis yielded only marginally different unstandardized regres- variance, F(5, 16) ⫽ 40.46, p ⬍ .001. As the moderators were only
sion coefficients, indicating that multicollinearity was not likely to be moderately correlated, we chose not to repeat the analysis with
a threat to the results.
error variance added in order to control for multicollinearity.

Restrictive Control
Inductive Control
The meta-analysis of the association between negative emotionality
and restrictive control included 22 studies and a total of 2,559 The meta-analysis of inductive control included 9 studies and a
mother– child dyads. In this set of studies, increased negative emo- total of 957 mother– child dyads. In this set of studies, higher
tionality was weakly but significantly associated with more restrictive levels of negative emotionality were weakly but significantly
control (r ⫽ .10, CI ⫽ .02–.17, random model, p ⬍ .01). The set of associated with less inductive control (r ⫽ .12, CI ⫽ .05–.18, fixed
448 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

Table 4
Moderators of Effect Sizes for Studies on Negative Emotionality and Restrictive Control

Moderator n k r 95% CI Qb Qw
**
Total set 2,559 22 .10 .02 to .17 69.10***
Publication yeara 0.18
Before 1987 519 7 .09 ⫺.10 to .27 22.52***
From 1987 on 2,040 15 .10* .01 to .19 45.40***
Impact factor 0.88
⬍2 1,633 13 .10* ⫺.01 to .21 47.81***
ⱖ2 926 9 .09 ⫺.02 to .20 20.41**
Type of study 2.61
Cross-sectional 2,363 20 .10** .02 to .18 66.04***
Longitudinal 196 2 .04 ⫺.11 to .18 0.44
Age of childb 26.32***
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Early infancy 603 5 .22*** .09 to .34 8.80


This document is copyrighted by the American Psychological Association or one of its allied publishers.

1-year-olds 603 8 ⫺.03 ⫺.15 to .10 15.08*


Preschoolers 1,353 9 .13** .04 to .23 18.89*
Gender 1.08
⬍50% boys 742 6 .14 ⫺.03 to .29 21.63***
ⱖ50% boys 1,717 15 .08 ⫺.02 to .18 44.72***
Percentage first born 13.28***
⬍75% 719 7 .19** .05 to .33 19.41**
75%–100% 332 4 ⫺.01 ⫺.12 to .10 0.30
Age of mother 2.15
⬍30 years 326 5 .08 ⫺.10 to .25 8.34
ⱖ30 years 762 7 .11 ⫺.03 to .25 19.30**
SES 2.07
Low–middle 1,923 15 .08* ⫺.01 to .18 49.05***
High 474 6 .08 ⫺.08 to .24 12.96*
Type of NE 9.90
Anger 876 4 .08 ⫺.16 to .31 22.04***
Fear 441 4 .06 ⫺.03 to .16 1.27
Anger and fear mixed 283 3 .04 ⫺.20 to .27 6.15*
Irritability 305 3 .14 ⫺.14 to .40 7.21*
Difficultness 614 7 .12 ⫺.05 to .29 22.52**
Negative mood 40 1 .00 ⫺.02 to .26 .00
Assessment method NE 10.33**
Parent report 1,472 12 .10* ⫺.01 to .20 34.04***
Observation 523 6 .02 ⫺.13 to .16 12.07*
Composite 564 4 .19 ⫺.01 to .36 12.65**
Assessment method parenting 12.07**
Parent report 1,080 5 .21*** .12 to .30 7.00
Observation 1,269 15 .06 ⫺.05 to .17 49.91***
Composite 210 2 .02 ⫺.12 to .16 0.12
Combination of assessment methods 21.16***
Parent report–parent report 1,080 5 .21*** .12 to .30 7.00
Observation–observation 392 5 .02 ⫺.17 to .20 11.93*
Parent report–observation 392 7 .00 ⫺.15 to .14 11.32
Composite 695 5 .15* ⫺.02 to .31 17.69**

Note. Unbiased effect sizes (rs) were calculated with one-tailed alpha set at .05. k ⫽ number of studies; CI ⫽
confidence interval; Qb ⫽ homogeneity statistic (between classes); Qw ⫽ homogeneity statistic (within classes).
a
Before 1987 ⫽ before Crockenberg’s (1986) review article; From 1987 on ⫽ after Crockenberg’s (1986)
original review article. b Early infancy ⫽ 3–11 months; 1-year-olds ⫽ 12–23 months; preschoolers ⫽ 24 –59
months.
*
p ⬍ .05. ** p ⬍ .01. *** p ⬍ .001.

model, p ⬍ .001). However, the calculated fail-safe number (k ⫽ Maternal Characteristics as Confounding Variables
9, fixed model) was 32, which is again below Rosenthal’s (1995)
critical value (9 ⫻ 5) ⫹ 10 ⫽ 55. This finding indicates that a As already noted in the introduction, if maternal characteristics
publication bias could exist, as only 32 studies reporting no asso- contribute to both ratings of children’s negative emotionality and
ciations between negative emotionality and inductive control were parenting behavior, the association between these two variables
needed to reduce this combined effect size to nonsignificance. The might be spurious because of inflated correlations. In addressing
set of effect sizes in this meta-analysis proved to be homogeneous, this issue, we started by investigating whether maternal character-
Qw(8) ⫽ 14.68, indicating that all variance has been explained, istics were indeed associated with both maternal ratings of nega-
making further analyses unnecessary. tive emotionality and observed or self-rated parenting.
CHILD NEGATIVE EMOTIONALITY AND PARENTING 449

Because only a small number of the selected studies provided associations) yielded larger effect sizes than studies in which
information about maternal characteristics, the three parenting maternal characteristics were only related to either negative emo-
categories were collapsed into one “optimal” parenting category, tionality or parenting (r ⬍ .10 for one of both associations). It
resulting in enhanced power. The maternal characteristics that turned out that nine of the selected studies provided information on
were reported predominantly concerned personality characteristics the association between maternal characteristics and mother-rated
(mostly derived from the Big Five or Big Three factor models) and temperament as well as on the association between maternal char-
personality-related constructs (such as depressive mood, stress, acteristics and parenting. The meta-analysis performed on this set
and self-efficacy) as well as some “support-related” maternal char- of studies yielded a combined effect size that just failed to reach
acteristics (e.g., low social support, low marital support). significance (r ⫽ .08, CI ⫽ ⫺.03 to ⫺.20, random model, p ⫽
The meta-analysis of the association between maternal charac- .07), with a heterogeneous set of unbiased effect sizes, Qw(8) ⫽
teristics and maternal ratings of negative emotionality included 12 20.27, p ⬍ .01. Moderator analyses revealed no differences in
studies and a total of 1,599 mother– child dyads. It showed unfa- effect sizes among studies in which maternal characteristics were
vorable maternal characteristics to be moderately associated with related to both mother-rated negative emotionality as well as
maternal ratings of negative emotionality (r ⫽ .25, CI ⫽ .16 –.34, parenting and studies in which one of these associations was very
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

random model, p ⬍ .001). The set of unbiased effect sizes proved small or absent, Qb(1) ⫽ 0.42, p ⫽ .52.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

to be heterogeneous, Qw(11) ⫽ 29.94, p ⬍ .01. The calculated


fail-safe number (k ⫽ 12) was 145 (random model), which is Discussion
above Rosenthal’s (1995) critical value of (12 ⫻ 5) ⫹ 10 ⫽ 70,
suggesting no publication bias. This meta-analytic review showed the combined effect size for
The meta-analysis of the association between maternal charac- the association between child negative emotionality and supportive
teristics and optimal parenting included 18 studies and a total of parenting as well as the combined effect size for the association
2,003 mother– child dyads. It showed unfavorable maternal char- between child negative emotionality and restrictive control to be
acteristics to be weakly associated with less optimal parenting (less small and to be moderated by sample and measurement character-
support, more restrictive control, and less inductive control; r ⫽ istics. The association between more negative emotionality and
.08, CI ⫽ .04 –.13, fixed model, p ⬍ .001). The set of unbiased less supportive parenting was relatively strong in families from
effect sizes proved to be homogeneous, Qw(17) ⫽ 19.81. The lower socioeconomic backgrounds, reversed in families from
calculated fail-safe number (k ⫽ 18) was 73 (fixed model), which higher socioeconomic backgrounds, and limited to studies with
is below Rosenthal’s (1995) critical value of (18 ⫻ 5) ⫹ 10 ⫽ 100, relatively high percentages of participants from ethnic minorities
indicating the possibility of publication bias. and studies using parent report to assess negative emotionality.
Having found associations between maternal characteristics and Higher levels of negative emotionality were associated with more
both negative emotionality and parenting, we proceeded in a restrictive control in samples with less than 75% first-born chil-
twofold way. First, we compared the effect sizes of studies con- dren as well as in infants and in preschoolers, but not in 1-year-
trolling for maternal characteristics with the effect sizes of the olds, and samples with predominantly first-born children. Further-
same studies when the results were not controlled for maternal more, only studies using parent report or composite measures to
characteristics. This was done by conducting two meta-analyses assess both negative emotionality and restrictive parenting yielded
using the same set of 5 studies (504 mother– child dyads). The significant effect sizes. The combined effect size for the relation
meta-analysis of uncontrolled associations between negative emo- between child negative emotionality and inductive control proved
tionality and parenting yielded a combined effect size of r ⫽ .18 to be small but homogeneous, so moderator analyses were not
for the fixed (CI ⫽ .09 –.26, p ⬍ .001) as well as the random (CI ⫽ performed for this meta-analysis. No differences in effect sizes
.06 –.30, p ⬍ .01) model. The set of unbiased effect sizes, Qw(4) ⫽ were found among studies reporting cross-sectional or longitudinal
6.46) was homogeneous. The calculated fail-safe number (k ⫽ 5) data in any of the three meta-analyses.
was 23 (fixed model), which is below Rosenthal’s (1995) critical The relation between negative emotionality and maternal sup-
value of (5 ⫻ 5) ⫹ 10 ⫽ 35, indicating the possibility of publi- port was found to be moderated by SES, ethnicity, and assessment
cation bias. The meta-analysis of associations controlled for ma- method of negative emotionality. The findings concerning SES
ternal characteristics still yielded a significant combined effect size support Crockenberg’s (1986) notion that mothers from resource-
(r ⫽ .15, CI ⫽ ⫺.01 to ⫺.30, random model, p ⬍ .05). The set of ful, middle-SES to high-SES backgrounds are capable of adapting
unbiased effect sizes, Qw(4) ⫽ 10.51, p ⬍ .05, in this meta- their parenting behavior to children who are harder to parent.
analysis was heterogeneous. The calculated fail-safe number was 2 Regarding ethnicity, it has been found that parenting in families
(random model), which is also below Rosenthal’s critical value of with an ethnic minority background, such as Hispanic and African
(5 ⫻ 5) ⫹ 10 ⫽ 35. Although the small fail-safe numbers indicate American, is usually less authoritative and more authoritarian
that these results should be interpreted with caution, these meta- (Brooks-Gunn & Markman, 2005). These families also value con-
analyses did not point into the direction of inflated correlations formity in their children more than Caucasian White families do
between negative emotionality and less optimal parenting, because (Ispa et al., 2004). Possibly, mothers from ethnic minorities are
the association between negative emotionality and parenting was less supportive when confronted with negative emotionality be-
still significant after controlling for maternal characteristics. cause they do not want to encourage their children’s undesirable
Second, we investigated possible spuriousness of the association behaviors and because they are less inclined to react from their
between negative emotionality and parenting by examining child’s point of view. Notably, the greater control exercised by
whether studies in which maternal characteristics were associated ethnic minority families may be adaptive in high-risk environ-
with parenting as well as negative emotionality (r ⱖ .10 for both ments (Deater-Deckard, Bates, Dodge, & Petit, 1996). Somewhat
450 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

unexpectedly, given the results for supportive parenting, the asso- reflects method variance. This may not be surprising, given the
ciation between negative emotionality and restrictive control finding that maternal characteristics were associated with both
proved not to be moderated by SES. A possible explanation for the mother-rated temperament as well as parenting. It is important that
absence of SES as a moderator of restrictive parenting may be method variance be accounted for in future studies by ensuring that
found in initial differences in the use of control strategies between measures of child temperament and maternal behavior are meth-
parents from lower and higher socioeconomic backgrounds. Ac- odologically independent, for example by observing parenting
cording to Hoff, Laursen, and Tardif (2002), it has been consis- behavior.
tently found that lower SES parents are more likely to use restric- Taken together, our findings illustrate that associations between
tive control strategies, including physical punishment, whereas negative emotionality and parenting are dependent on context and
higher SES parents are more likely to use inductive control strat- measurement characteristics. In line with this assertion, Crocken-
egies, including reasoning and appeals to guilt. It seems that lower berg and Leerkes (2003) suggested that the type of association
SES parents with highly negative emotional children do not become between negative emotionality and parenting might vary between
more restrictive in their control strategies than they already are. high-risk and low-risk samples, with main effects more likely to be
Mothers were found to use more restrictive control toward found in high-risk samples, because protective factors are less
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

infants and preschoolers high in negative emotionality but not prevalent there. The findings of our meta-analytic review particu-
toward 1-year-olds. This is not in line with results from, for larly underscore the importance of taking into account maternal
example, Lee and Bates (1985), showing that over time mothers characteristics in research on temperament and parenting, as this
responded more intrusively to children high in negative emotion- may help to clarify the issue of direction of effects. Studies using
ality. A somewhat tentative explanation for our finding might be large samples of families that are tracked over a longer period of
that children differ in the type of exploratory behavior they show time can still be considered to provide the most compelling data on
at different ages. Compared with infants and preschoolers, children the mutual impact of infants and caregivers on each other, espe-
in the 2nd year of life can be considered relatively vigorous cially when data obtained from the prospective parents are in-
explorers of their environment. Possibly, in order to prevent chil- cluded as control variables (Crockenberg & Leerkes, 2003). How-
dren from damaging material goods and because children might ever, in cases in which this cannot be attained, these authors
easily hurt themselves, parents of children in the 2nd year of life suggest that researchers might add other methodologies, such as
elicit a constantly high level of restrictive control, resulting in the microlevel approaches, growth-curve modeling (e.g., van den
absence of an association between negative emotionality and par- Boom & Hoeksma, 1995), or qualitative data, to their correlational
enting during this age range. If this explanation, together with the approach.
explanation in the previous paragraph, holds, this should have Several limitations of the present meta-analytic review should
consequences not only for the measurement of restrictive control be mentioned. To begin with, our results should be interpreted with
but also for the measurement of other parenting behaviors across great care, as many studies that were used in our meta-analyses
ages and situations. All types of parenting behavior measured may suffer from considerable methodological drawbacks, such as lack
tap into different constructs at different developmental stages and of independence between measures of negative emotionality and
in different social contexts. parenting, possible influences of child-rearing history on concur-
Although more negative emotionality proved to be weakly as- rent associations between child temperament and parenting, inabil-
sociated with inductive control, this finding was straightforward in ity to account for genetic effects, and reliance on correlational data
that the combined effect size proved to be homogeneous. Homo- (Putnam et al., 2002; Sanson et al., 2004). Therefore, conclusions
geneity of effect sizes in the domain of inductive control may be about the causal relation between temperament and parenting
explained by the fact that eight of the nine studies in this meta- cannot be drawn.
analysis were conducted in samples with predominantly Caucasian A further limitation of this meta-analytic review is the possibil-
White mother– child dyads from middle to high socioeconomic ity of file drawer problems, indicating publication bias. The po-
backgrounds. tential publication bias found in the meta-analysis has important
The suggestion that maternal characteristics would be associated implications for theoretical models of the association between
with both mother-rated temperament as well as parenting was children’s negative emotionality and parenting, for instance, be-
supported in this meta-analytic review. However, controlling for cause it could hamper the generalizability of findings. As a con-
maternal characteristics still yielded a significant association be- sequence, our results have to be interpreted with caution. It must
tween negative emotionality and parenting, which suggests that be noted, however, that the fail-safe number, as a test of possible
our results concerning the association between negative emotion- publication bias, is related to the magnitude of the effect sizes
ality and parenting do not have to be considered as spurious. (Rosenberg, 2005). As effect sizes become larger, the fail-safe
Moreover, we found no differences in effect sizes among studies in number rapidly increases to exceed Rosenthal’s (2005) critical
which maternal characteristics were related to both mother-rated value.
negative emotionality as well as parenting and studies in which Finally, it can be considered a limitation of the field in general
one of these associations was very small or absent. that studies focus almost exclusively on mothers. A more complete
The finding that significant associations between negative emo- examination of the association between negative emotionality and
tionality and supportive parenting were limited to studies using parenting requires a greater number of studies that include data
parent report of negative emotionality as well as the finding that from fathers. Unfortunately, we only encountered four studies in
significant associations between negative emotionality and restric- which fathers were involved in one way or another. Only two of
tive parenting were limited to studies using parent report or com- these studies separately reported an association between negative
posite measures of negative emotionality and parenting probably emotionality and parenting for fathers. Worobey et al. (1983)
CHILD NEGATIVE EMOTIONALITY AND PARENTING 451

found fathers and mothers of 1-month-old irritable infants to show Brooks-Gunn, J., & Markman, L. B. (2005). The contribution of parenting
similar soothing behaviors. In addition, Belsky et al. (1998) found to ethnic and racial gaps in school readiness. Future of Children, 15,
no differences in mothering and fathering as a function of child 139 –168.
negativity. Calkins, S. D. (2002). Does aversive behavior during toddlerhood matter?
On the whole, studying the link between negative emotionality The effects of difficult temperament on maternal perceptions and be-
havior. Infant Mental Health Journal, 23, 381– 402.
and parenting can be compared with aiming at a moving target.
*Calkins, S. D., Hungerford, A., & Dedmon, S. E. (2004). Mother’s
From the moment that parent– child interaction develops, the ex- interactions with temperamentally frustrated infants. Infant Mental
pression of negative emotionality is influenced by parental behav- Health Journal, 25, 219 –239.
ior. Parental behavior, in turn, is influenced by parental percep- *Calkins, S. D., & Johnson, M. C. (1998). Toddler regulation of distress to
tions of the child and attributions about parenting the child. frustrating events: Temperamental and maternal correlates. Infant Be-
Parental behavior as well as parental attributions and perceptions havior and Development, 21, 379 –395.
are in turn influenced by parental history, the social and physical *Campbell, S. B. (1979). Mother–infant interaction as a function of ma-
context of the family, and other characteristics of the child. Re- ternal ratings of temperament. Child Psychiatry and Human Develop-
search into negative emotionality and parenting is additionally ment, 10, 67–76.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

Campbell, S. B. (1995). Behavior problems in preschool children: A


This document is copyrighted by the American Psychological Association or one of its allied publishers.

complicated by the fact that different sources of information about


children’s temperament (e.g., parents, teachers, researchers) prob- review of recent research. Journal of Child Psychology and Psychiatry
and Allied Disciplines, 36, 113–149.
ably tap different components of the construct.
Chess, S., & Thomas, A. (1984). Origins and evolution of behavior
Notwithstanding the methodological drawbacks that character-
disorders: From infancy to early adult life. New York: Brunner/Mazel.
ize research on temperament and parenting, this meta-analytic *Clark, L. A., Kochanska, G., & Ready, R. (2000). Mothers’ personality
review shows negative emotionality to be a concept helpful in and its interaction with child temperament as predictors of parenting
explaining patterns of parenting in a diversity of settings, thereby behavior. Journal of Personality and Social Psychology, 79, 274 –285.
providing insight into the complexity of the association between *Clark, R., Hyde, J. S., Essex, M. J., & Klein, M. H. (1997). Length of
negative emotionality and parenting. The results of this meta- maternity leave and quality of mother–infant interactions. Child Devel-
analytic review are supportive of Crockenberg’s (1986) suggestion opment, 68, 364 –383.
that relations between negative emotionality and parenting are *Clarke-Stewart, K. A. (1973). Interactions between mothers and their
influenced by characteristics of the caregiver or the caregiving young children: Characteristics and consequences. Monographs of the
environment, and they underscore the importance of maternal Society for Research in Child Development, 38(6 –7, Serial No. 153).
perceptions of child negative emotionality for the study of tem- *Coffman, S., Levitt, M. J., Guacci, N., & Silver, M. (1992). Temperament
and interactive effects: Mothers and infants in a teaching situation.
perament and parenting.
Issues in Comprehensive Pediatric Nursing, 15, 169 –182.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences.
References Hillsdale, NJ: Erlbaum.
References marked with an asterisk indicate studies included in the Coplan, R. J., Hastings, P. D., Lagace-Seguin, D. G., & Moulton, C. E.
meta-analysis. (2002). Authoritative and authoritarian mothers’ parenting goals, attri-
Bates, J. E. (1989). Concepts and measures of temperament. In G. A. butions and emotions across different childrearing contexts. Parenting:
Kohnstamm & J. E. Bates (Eds.), Temperament in childhood (pp. 3–26). Science and Practice, 2, 1–26.
New York: Wiley. *Coplan, R. J., Prakash, K., O’Neil, K., & Armer, M. (2004). Do you
*Bates, J. E., Olson, S. L., Pettit, G. S., & Bayles, K. (1982). Dimensions “want” to play? Distinguishing between conflicted shyness and social
of individuality in the mother–infant relationship at six months of age. disinterest in early childhood. Developmental Psychology, 40, 244 –258.
Child Development, 53, 446 – 461. Crockenberg, S. (1986). Are temperamental differences in babies associ-
Bates, J. E., Pettit, G. S., Dodge, K. A., & Ridge, B. (1998). Interaction of ated with predictable differences in care-giving? In J. Lerner & R.
temperamental resistance to control and restrictive parenting in the Lerner (Eds.), Temperament and psychosocial interaction in children:
development of externalizing behavior. Developmental Psychology, 34, New directions for child development (Vol. 31, pp. 53–73). San Fran-
982–995. cisco: Jossey-Bass.
Belsky, J. (1984). The determinants of parenting: A process model. Child Crockenberg, S., & Leerkes, E. (2003). Infant negative emotionality,
Development, 55, 83–96. caregiving, and family relationships. In A. Booth & A. C. Crouter (Eds.),
*Belsky, J., Hsieh, K. H., & Crnic, K. (1998). Mothering, fathering, and Children’s influence on family dynamics: The neglected side of family
infant negativity as antecedents of boys’ externalizing problems and relationships (pp. 57–78). Mahwah, NJ: Erlbaum.
inhibition at age 3 years: Differential susceptibility to rearing experi- *Crockenberg, S. B., & Smith, P. (1982). Antecedents of mother–infant
ence? Development and Psychopathology, 10, 301–319. interaction and infant irritability in the first 3 months of life. Infant
Bernaards, C. A., & Sijtsma, K. (1999). Factor analysis of polytomous item Behavior and Development, 5, 105–119.
response data suffering from ignorable item nonresponse. Multivariate *Daniels, D., Plomin, R., & Greenhalgh, J. (1984). Correlates of difficult
Behavioral Research, 34, 277–313. temperament in infancy. Child Development, 55, 1184 –1194.
Bornstein, M. H., Hahn, C.-S., Suwalsky, J. T. D., & Haynes, O. M. (2003). Deater-Deckard, K., Bates, J. E., Dodge, K. A., & Petit, G. S. (1996).
Socioeconomic status, parenting and child development: The Hollings- Physical discipline among African American and European American
head Four-Factor Index of Social Status and the Socioeconomic Index of mothers: Links to children’s externalizing behaviors. Developmental
Occupations. In M. H. Bornstein & R. H. Bradley (Eds.), Socioeconomic Psychology, 32, 1065–1072.
status, parenting, and child development (pp. 29 – 82). Mahwah, NJ: *Dunn, J., & Kendrick, C. (1980). Studying temperament and parent– child
Erlbaum. interaction: Comparison of interview and direct observation. Develop-
Borenstein, M., Rothstein, D., & Cohen, J. (2000). Comprehensive meta- mental Medicine and Child Neurology, 22, 484 – 496.
analysis: A computer program for research synthesis [Computer soft- Durlak, J. A., & Lipsey, M. W. (1991). A practitioner’s guide to meta-
ware manual]. Englewood, NJ: Biostat. analysis. American Journal of Community Psychology, 19, 291–333.
452 PAULUSSEN-HOOGEBOOM, STAMS, HERMANNS, AND PEETSMA

*Eisenberg, N., & Fabes, R. A. (1994). Mothers’ reactions to children’s security of attachment: Multiple pathways to emerging internalization.
negative emotions: Relations to children’s temperament and anger be- Child Development, 66, 597– 615.
havior. Merrill-Palmer Quarterly, 40, 138 –156. *Kochanska, G. (1998). Mother– child relationship, child fearfulness, and
*Fish, M., & Stifter, C. A. (1993). Mother parity as a main and moderating emerging attachment: A short-term longitudinal study. Developmental
influence on early mother–infant interaction. Journal of Applied Devel- Psychology, 34, 480 – 490.
opmental Psychology, 14, 557–572. *Kochanska, G., Friesenborg, A. E., Lange, L. A., & Martel, M. M. (2004).
*Fish, M., Stifter, C. A., & Belsky, J. (1991). Conditions of continuity and Parents’ personality and infants’ temperament as contributors to their
discontinuity in infant negative emotionality: Newborn to five months. emerging relationship. Journal of Personality and Social Psychology,
Child Development, 62, 1525–1537. 86, 744 –759.
*Gauvain, M., & Fagot, B. (1995). Child temperament as a mediator of *Kyrios, M., & Prior, M. (1990). Temperament, stress, and family factors
mother–toddler problem solving. Social Development, 4, 257–276. in behavioural adjustment of 3–5 year-old children. International Jour-
*Gordon, B. N. (1983). Maternal perception of child temperament and nal of Behavioral Development, 13, 67–93.
observed mother– child interaction. Child Psychiatry and Human Devel- Lee, C. L., & Bates, J. E. (1985). Mother– child interaction at age two years
opment, 13, 153–167. and perceived difficult temperament. Child Development, 56, 1314 –
1325.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

*Hagekull, B., & Bohlin, G. (1986). Mother–infant interaction and per-


*Leerkes, E. M., & Crockenberg, S. C. (2003). The impact of maternal
This document is copyrighted by the American Psychological Association or one of its allied publishers.

ceived temperament. International Journal of Behavioral Development,


6, 297–313. characteristics and sensitivity on the concordance between maternal
*Hagekull, B., Bohlin, G., & Rydell, A. M. (1997). Maternal sensitivity, reports and laboratory observations of infant negative emotionality.
infant temperament, and the development of early feeding problems. Infancy, 4, 517–539.
Infant Mental Health Journal, 18, 92–106. *Lehman, E. B., Steier, A. J., Guidash, K. M., & Wanna, S. Y. (2002).
*Halpern, L. F., & McLean, W. E., Jr. (1997). “Hey mom, look at me!” Predictors of compliance in toddlers: Child temperament, maternal per-
Infant Behavior and Development, 20, 515–529. sonality, and emotional availability. Early Child Development and Care,
Hampel, F. R., Ronchetti, E. M., Rousseeuw, P. J., & Stahel, W. A. (1986). 173, 301–310.
Robust statistics: The approach based on influence functions. New *Lerner, J. V., & Galambos, N. L. (1985). Maternal role satisfaction,
mother– child interaction, and child temperament: A process model.
York: Wiley.
Developmental Psychology, 21, 1157–1164.
*Hann, D. M. (1989). A systems conceptualization of the quality of
Lerner, R. M., Rothbaum, F., Boulos, S., & Castellino, D. R. (2002).
mother–infant interaction. Infant Behavior and Development, 12, 251–
Developmental systems perspective on parenting. In M. H. Bornstein
263.
(Ed.), Handbook of parenting: Vol. 2. Biology and ecology of parenting
*Harrington, D., Black, M. M., Starr, R. H., Jr., & Dubowitz, H. (1998).
(pp. 315–344). Mahwah, NJ: Erlbaum.
Child neglect: Relation to child temperament and family context. Amer-
Lipsey, M. W., & Wilson, D. B. (2000). Practical meta-analysis (Vol. 49).
ican Journal of Orthopsychiatry, 68, 108 –116.
London: Sage.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis.
*Luster, T., Boger, R., & Hannan, K. (1993). Infant affect and home
San Diego, CA: Academic Press.
environment. Journal of Marriage and the Family, 55, 651– 661.
Hoff, E., Laursen, B., & Tardif, T. (2002). Socioeconomic status and
Maccoby, E. E., & Martin, J. A. (1983). Socialization in the context of the
parenting. In M. H. Bornstein (Ed.), Handbook of parenting (Vol. 2, pp.
family: Parent– child interaction. In P. H. Mussen (Ed.), Handbook of
231–252). Mahwah, NJ: Erlbaum.
child psychology: Vol. 4. Socialization, personality, and social develop-
Hollingshead, A. B. (1975). Four Factor Index of Social Status. New
ment (pp. 1–101). New York: Wiley.
Haven, CT: Yale University.
*Maccoby, E. E., Snow, M. E., & Jacklin, C. N. (1984). Children’s
Hollingshead, A. B., & Redlich, F. C. (1958). Social class and mental dispositions and mother– child interaction at 12 and 18 months: A
illness. New York: Wiley. short-term longitudinal study. Developmental Psychology, 20, 459 – 472.
*Houldin, A. D. (1987). Infant temperament and the quality of the child- *Mangelsdorf, S. C., Gunnar, M., Kestenbaum, R., Lang, S., & Andreas, D.
rearing environment. Maternal-Child Nursing Journal, 16, 131–143. (1990). Infant proneness-to-distress temperament, maternal personality,
Ispa, J. M., Fine, M. A., Halgunseth, L. C., Harper, S., Robinson, J., Boyce, and mother–infant attachment: Associations and goodness of fit. Child
L., et al. (2004). Maternal intrusiveness, maternal warmth, and mother– Development, 61, 820 – 831.
toddler relationship outcomes: Variations across low-income ethnic and *Mangelsdorf, S. C., McHale, J. L., Diener, M., Heim Goldstein, L., &
acculturation groups. Child Development, 75, 1613–1631. Lehn, L. (2000). Infant attachment: Contributions of infant temperament
*Karraker, K. H., Lake, M. A., & Parry, T. B. (1994). Infant coping with and maternal characteristics. Infant Behavior and Development, 23,
everyday stressful events. Merrill-Palmer Quarterly, 40, 171–189. 175–196.
*Katainen, S., Räikkönen, K., & Keltikangas-Järvinen, L. (1997). Child- *Maziade, M., Cote, R., Bernier, H., Boutin, P., & Thivierge, J. (1989).
hood temperament and mother’s child-rearing attitudes: Stability and Significance of extreme temperament in infancy for clinical status in
interaction in a three-year follow-up study. European Journal of Per- preschool years: I. Value of extreme temperament at 4 – 8 months for
sonality, 11, 249 –265. predicting diagnosis at 4.7 years. British Journal of Psychiatry, 154,
*Keefe, M., Kotzer, A. M., Froese-Fretz, A., & Curtin, M. (1996). A 535–543.
longitudinal comparison of irritable and nonirritable infants. Nursing *Mertesacker, B., Bade, U., Haverkock, A., & Pauli-Pott, U. (2004).
Research, 45, 4 –9. Predicting maternal reactivity/sensitivity: The role of infant emotional-
Kerr, D. C. R., Lopez, N. L., Olson, S. L., & Sameroff, A. J. (2004). ity, maternal depressiveness/anxiety, and social support. Infant Mental
Parental discipline and externalizing behavior problems in early child- Health Journal, 25, 47– 61.
hood: The roles of moral regulation and child gender. Journal of Ab- *Milliones, J. (1978). Relationship between perceived child temperament
normal Child Psychology, 32, 369 –383. and maternal behaviors. Child Development, 49, 1255–1257.
*Kiang, L., Moreno, A. J., & Robinson, J. L. (2004). Maternal preconcep- *Moss, H. A. (1967). Sex, age, and state as determinants of mother–infant
tions about parenting predict child temperament, maternal sensitivity, interaction. Merrill-Palmer Quarterly, 13, 19 –36.
and children’s empathy. Developmental Psychology, 40, 1081–1092. Mullen, B. (1989). Advanced basic meta-analysis. Hillsdale, NJ: Erlbaum.
*Kochanska, G. (1995). Children’s temperament, mothers’ discipline, and *Nachmias, M., Gunnar, M., Mangelsdorf, S., Parritz, R. H., & Buss, K.
CHILD NEGATIVE EMOTIONALITY AND PARENTING 453

(1996). Behavioral inhibition and stress reactivity: The moderating role Shiner, R. L. (1998). How shall we speak of children’s personalities in
of attachment security. Child Development, 67, 508 –522. middle childhood? A preliminary taxonomy. Psychological Bulletin,
*NICHD Early Child Care Research Network. (1999). Child care and 124, 308 –332.
mother– child interaction in the first three years of life. Developmental *Simonds, M. P., & Simonds, J. F. (1981). Relationship of maternal
Psychology, 35, 1399 –1413. parenting behaviors to preschool children’s temperament. Child Psychi-
*Owens, E. B., Shaw, D. S., & Vondra, J. I. (1998). Relations between atry and Human Development, 12, 19 –31.
infant irritability and maternal responsiveness in low-income families. *Simpson, A. E., & Stevenson-Hinde, J. (1985). Temperamental charac-
Infant Behavior and Development, 21, 761–778. teristics of three- to four-year-old boys and girls and child–family
*Pauli-Pott, U., Mertesacker, B., Bade, U., Bauer, C., & Beckmann, D. interactions. Journal of Child Psychology and Psychiatry, 26, 43–53.
(2000). Contexts of relations of infant negative emotionality to caregiv- *Spangler, G. (1990). Mother, child, and situational correlates of toddler’s
ers reactivity/sensitivity. Infant Behavior and Development, 23, 23–39. social competence. Infant Behavior and Development, 13, 405– 419.
*Peters-Martin, P., & Wachs, T. D. (1984). A longitudinal study of *Susman-Stillman, A., Kalkoske, M., Egeland, B., & Waldman, I. (1996).
temperament and its correlates in the first 12 months. Infant Behavior Infant temperament and maternal sensitivity as predictors of attachment
and Development, 7, 285–298. security. Infant Behavior and Development, 19, 33– 47.
Prior, M. (1992). Childhood temperament. Journal of Child Psychology Tabachnick, B. G., & Fidell, L. S. (2001). Using multivariate statistics.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.

and Psychiatry and Allied Disciplines, 33, 249 –279. New York: Harper & Row.
This document is copyrighted by the American Psychological Association or one of its allied publishers.

*Prior, M., Sanson, A., Carroll, R., & Oberklaid, F. (1989). Social class Thomas, A., & Chess, S. (1977). Temperament and development. New
differences in temperament ratings by mothers of preschool children. York: Brunner/Mazel.
Merrill-Palmer Quarterly, 35, 239 –248. Thomas, A., Chess, S., & Birch, H. G. (1968). Temperament and behavior
Putnam, S. P., Sanson, A. V., & Rothbart, M. K. (2002). Child tempera- disorders in children. New York: New York University Press.
ment and parenting. In M. H. Bornstein (Ed.), Handbook of parenting Thomas, A., Chess, S., Birch, H. G., Hertzig, M. E., & Korn, S. (1963).
(Vol. 1, pp. 255–277). London: Erlbaum. Behavioural individuality in early childhood. New York: New York
Rollins, B. C., & Thomas, D. L. (1979). Parental support, power, and University Press.
control techniques in the socialization of children. In W. R. Burr, R. Hill, Thomson Corporation. (2003). ISI journal citation reports, science edition.
F. I. Nye, & I. L. Reiss (Eds.), Contemporary theories about the family Retrieved June, 16, 2005, from http://isi15.isiknowledge.com/portal.cgi/jcr
(Vol. 1, pp. 317–364). London: Free Press. van den Boom, D. C., & Hoeksma, J. B. (1995). The effect of infant
Rosenberg, M. S. (2005). The file-drawer problem revisited: A general irritability on mother–infant interaction: A growth-curve analysis. De-
weighted method for calculating fail-safe numbers in meta-analysis. velopmental Psychology, 30, 581–590.
Evolution, 59, 464 – 468. *van der Mark, I. L., Bakermans-Kranenburg, M. J., & van IJzendoorn,
Rosenthal, R. (1995). Writing meta-analytic reviews. Psychological Bul- M. H. (2002). The role of parenting, attachment, and temperamental
letin, 118, 183–192. fearfulness in the prediction of compliance in toddler girls. British
*Rothbart, M. K. (1986). Longitudinal observation of infant temperament. Journal of Developmental Psychology, 20, 361–378.
Developmental Psychology, 22, 356 –365. van IJzendoorn, M. H., & Bakermans-Kranenburg, M. J. (2004). Maternal
Rothbart, M. K., Ahadi, S. A., & Hershey, K. L. (1994). Temperament and sensitivity and infant temperament in the formation of attachment. In G.
social behavior in childhood. Merrill-Palmer Quarterly, 40, 21–39. Bremner & A. Slater (Eds.), Theories of infant development (pp. 233–
Rothbart, M. K., & Bates, J. E. (1998). Temperament. In W. Damon & N. 257). Oxford, England: Blackwell.
Eisenberg (Eds.), Handbook of child psychology: Social, emotional and van IJzendoorn, M. H., Vereijken, C. M. J. L., Bakermans-Kranenburg,
personality development (5th ed., Vol. 3, pp. 105–176). New York: M. J., & Riksen-Walraven, J. M. (2004). Assessing attachment security
Wiley. with the Attachment Q Sort: Meta-analytic evidence for the validity of
Rothbaum, R., & Weisz, J. R. (1994). Parental caregiving and child the observer AQS. Child Development, 75, 1188 –1213.
externalizing behavior in nonclinical samples: A meta-analysis. Psycho- *Vaughn, B. E., Taraldson, B. J., Crichton, L., & Egeland, B. (1981). The
logical Bulletin, 116, 55–74. assessment of infant temperament: A critique of the Carey Infant Tem-
*Rubin, K. H., Burgess, K. B., & Hastings, P. D. (2002). Stability and perament Questionnaire. Infant Behavior and Development, 4, 1–17.
social– behavioral consequences of toddlers’ inhibited temperament and Washington, J., Minde, K., & Goldberg, S. (1986). Temperament in pre-
parenting behaviors. Child Development, 73, 483– 495. term infants: Style and stability. Journal of the American Academy of
Rubin, K. H., Stewart, S. L., & Chen, X. (1995). Parents of aggressive and Child and Adolescent Psychiatry, 25, 493–502.
withdrawn children. In M. Borenstein (Ed.), Handbook of parenting: *Worobey, J., Laub, K. W., & Schilmoeller, G. L. (1983). Maternal and
Vol. 1. Children and parenting (pp. 255–284). Hillsdale, NJ: Erlbaum. paternal responses to infant distress. Merrill-Palmer Quarterly, 29, 33– 45.
Sanson, A., Hemphill, S. A., & Smart, D. (2004). Connections between Wright Guerin, D., Gottfried, A. W., & Thomas, C. W. (1997). Difficult
temperament and social development: A review. Social Development, temperament and behaviour problems: A longitudinal study from 1.5 to
13, 142–170. 12 years. International Journal of Behavioral Development, 21, 71–90.
*Schuler, M. E., Black, M. M., & Starr, R. H., Jr. (1995). Determinants of Zahr, L. (1991). Correlates of mother–infant interaction in premature
mother–infant interaction: Effects of prenatal drug exposure, social infants from low socioeconomic backgrounds. Pediatric Nursing, 17,
support, and infant temperament. Journal of Clinical Child Psychology, 259 –264.
24, 397– 405. Zeanah, C. H., & Fox, N. A. (2004). Temperament and attachment disor-
*Seifer, R., Schiller, M., Sameroff, A. J., Resnick, S., & Riordan, K. ders. Journal of Clinical Child and Adolescent Psychology, 33, 32– 41.
(1996). Attachment, maternal sensitivity, and infant temperament during
the first year of life. Developmental Psychology, 32, 12–25.
*Shaw, D. S., & Vondra, J. I. (1995). Infant attachment security and maternal Received July 6, 2005
predictors of early behavior problems: A longitudinal study of low-income Revision received July 17, 2006
families. Journal of Abnormal Child Psychology, 23, 335–357. Accepted August 1, 2006 䡲

You might also like