Is the ideal mother a sensitive mother?
Beliefs about early childhood parenting
in mothers across the globe
Judi Mesman,
1
Marinus van IJzendoorn,
1
Kazuko Behrens,
2
Olga Alicia Carbonell,
3
Rodrigo Ca
´
rcamo,
1,4
Inbar Cohen-Paraira,
5
Christian de la Harpe,
6
Hatice Ekmekc¸i,
1
Rosanneke Emmen,
1
Jailan Heidar,
1
Kiyomi Kondo-Ikemura,
7
Cindy Mels,
8
Haatembo Mooya,
9
Sylvia Murtisari,
10
Magaly No
´
blega,
11
Jenny Amanda Ortiz,
12
Abraham Sagi-Schwartz,
5
Francis Sichimba,
9
Isabel Soares,
13
Howard Steele,
14
Miriam Steele,
14
Marloes Pape,
1
Joost van Ginkel,
1
Rene
´
van der Veer,
1
Lamei Wang,
15
Bilge Selcuk,
16
Melis Yavuz,
16
and Ghadir Zreik
5
Abstract
In this article, we test the hypothesis that beliefs about the ideal mother are convergent across cultures and that these beliefs overlap
considerably with attachment theory’s notion of the sensitive mother. In a sample including 26 cultural groups from 15 countries
around the globe, 751 mothers sorted the Maternal Behavior Q-Set to reflect their ideas about the ideal mother. The results show
strong convergence between maternal beliefs about the ideal mother and attachment theory’s description of the sensitive mother
across groups. Cultural group membership significantly predicted variations in maternal sensitivity belief scores, but this effect was
substantially accounted for by group variations in socio-demographic factors. Mothers living in rural versus urban areas, with a low
family income, and with more children, were less likely to describe the ideal mother as highly sensitive. Cultural group membership did
remain a significant predictor of variations in maternal sensitivity belief scores above and beyond socio-demographic predictors. The
findings are discussed in terms of the universal and culture-specific aspects of the sensitivity construct.
Keywords
cross-culture, maternal sensitivity, mother–infant relationships, socioeconomic status
Attachment theory was formulated to represent a universally applica-
ble account of the bond between caregivers and infants based
on evolutionary and ethological considerations (Bowlby, 1969).
Although the number of cross-cultural studies is still limited, empiri-
cal research indeed provides some support for the universality of the
major tenets of attachment theory, with evidence for the universality
hypothesis that across the world, (virtually) all infants become
attached to one or more specific caregivers, and the normativity
hypothesis that secure attachment is the most common form of
attachment across cultures (Mesman, Van IJzendoorn, & Sagi-
Schwartz, in press). In addition, there is evidence that maternal
beliefs about the ideal child overlap considerably with the notion
of secure-base behavior and show high agreement across cultures
(Posada et al., 1995, 2013; see also Sternberg & Lamb, 1992).
Cross-cultural research on the tenets of attachment theory has mostly
focused on child behaviors, and less attention has been paid to the
parental side of the attachment coin. Sensitive parenting is defined
as a caregiver’s ability to perceive child signals, to interpret these sig-
nals correctly, and to respond to them contingently and appropriately
(Ainsworth, Bell, & Stayton, 1974), and has been clearly identified
1
Centre for Child and Family Studies, Leiden University, the Netherlands
2
State University of New York, Albany, NY, USA
3
Pontificia Universidad Javeriana, Bogota, Colombia
4
University of Magallanes, Punta Arenas, Chile
5
University of Haifa, Israel
6
Catholic University of Temuco, Temuco, Araucania, Chile
7
Tokyo University, Japan
8
Catholic University of Uruguay, Montevideo, Uruguay
9
University of Zambia, Lusaka, Zambia
10
Sanata Dharma University, Daerah Istimewa Yogyakarta, Indonesia
11
Pontificia Universidad Cato
´
lica del Peru
´
, Lima, Peru
12
Universidade Federal do Rio Grande do Sul, Porto Alegre – RS, Brazil
13
University of Minho, Guimaraes, Portugal
14
The New School for Social Research, New York, NY, USA
15
Chinese Academy of Sciences, Beijing, China
16
Koc University, Sarıyer/
_
Istanbul, Turkey
Corresponding author:
Judi Mesman, Centre for Child and Family Studies, Leiden University,
Wassenaarseweg 52, 2333 AK, Leiden, the Netherlands.
Email: mesmanj@fsw.leidenuniv.nl
International Journal of
Behavioral Development
1–13
ª The Author(s) 2015
Reprints and permissions:
sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0165025415594030
ijbd.sagepub.com
at Universiteit Leiden \ LUMC on November 21, 2016jbd.sagepub.comDownloaded from
as an important parental characteristic that facilitates secure attach-
ment in children (Bakermans-Kranenburg, Van IJzendoorn, & Juffer,
2003). In the current study that includes 751 mothers from 26 cultural
groups in 15 countries, we test the hypothesis that maternal beliefs
about the ideal mother converge considerably with the notion of the
highly sensitive mother, and we test socio-demographic and cultural
predictors of individual variations in this convergence.
Culture can be broadly defined as patterns of behaviors, knowl-
edge, and beliefs acquired through socialization processes, and
that distinguishes one group from another group (e.g., Boyd &
Richerson, 2005). Two of the most commonly-used distinctions in
the literature refer to individualism versus collectivism (Oyserman,
Coon, & Kemmelmeier, 2002), and autonomous versus relatedness
cultural orientation (Kagitcibasi, 2007). In individualistic and auton-
omous cultures (generally found in urban Western areas), the individ-
ual is valued over the group, and parenting is geared towards
fostering psychological independence and individual achievement,
and is characterized by a distal parenting style and face-to-face con-
tact rather than physical proximity (Feldman & Masalha, 2010; Kel-
ler et al., 2009). In collectivistic cultures and cultures emphasizing
relatedness (generally found in non-Western rural areas), the group
is valued over the individual, and parenting focuses on fostering
social relations and hierarchies and stimulating obedience to group
norms, and is characterized by a proximal parenting style expressed
by physical closeness (e.g., Keller et al., 2009). A blend of these two
broad cultural orientations is described by the autonomous-related
cultural model, which characterizes urbanized or migrated groups
that originally come from collectivistic cultural backgrounds, but
now reside in a more individualistic context (Kagitcibasi, 2007).
It has been argued that the definitions of the main attachment-
related constructs, including secure attachment and sensitivity,
are biased towards individualistic cultural contexts with distal par-
enting patterns because of their focus on meeting the individual
infant’s needs and fostering psychological autonomy (e.g., Keller,
2013; Rothbaum, Weisz, Pott, Miyake, & Morelli, 2000). However,
Mary Ainsworth’s first descriptions of behaviors that reflect sensi-
tive responsiveness were based on her observations in Uganda
(Ainsworth, 1967). Cross-cultural research indeed confirms that
sensitivity and attachment can be validly measured in non-
Western contexts, and also relates to secure infant attachment in
the expected directions in countries such as China (Ding, Xiu,
Wang, Li, & Wang, 2012), Japan (Vereijken, Riksen-Walraven,
& Kondo-Ikemura, 1997), South Korea (Jin, Jacobvitz, Hazen, &
Jung, 2012), Mali (True, Pisani, & Oumar, 2001), Mexico (Gojman
et al., 2012), and South Africa (Tomlinson, Cooper, & Murray,
2005). Further, some of the main components of sensitivity repre-
sent universally important aspects of caregiving. Availability and
proximity are crucial to infant survival in that the child is kept safe.
Prompt responding serves the child’s common human ability to
detect contingencies between one’s own behavior and environmen-
tal events, and child-centered responsiveness is further conducive
of child well-being in that it will get fed when signaling hunger,
protected when signaling fear, and cared for when signaling pain.
One way of addressing the universality versus culture-specificity
of core attachment concepts is to examine parental beliefs (or
ethnotheories) about these concepts in different cultures. Posada
and colleagues (1995) used the Attachment Q-Set (AQS) to assess
maternal beliefs about ideal child behavior and found that these
were similar across seven Western and non-Western countries and
overlapped considerably with attachment theory’s notion of the
secure-base phenomenon (i.e., a child staying close to caregiver
in unfamiliar situations, easily comforted by caregiver, and explora-
tive when put at ease). In a recent study (Emmen, Malda, Mesman,
Ekmekci, & Van IJzendoorn, 2012), the Maternal Behavior Q-Set
(MBQS; Pederson & Moran, 1995) was used to assess maternal
beliefs about the ideal mother in relation to the sensitivity construct
in three different ethnic groups within the Netherlands (majority
Dutch, minority Turkish and Moroccan with an immigrant back-
ground). The results showed strong convergence between maternal
views of the ideal mother and attachment theory’s notion of sensi-
tive parenting across cultural groups (Emmen et al., 2012). How-
ever, the cultural groups in this study all resided within the same
country, and the minority groups represented mostly second-
generation immigrants who were born in the Netherlands. Thus,
replication and extension of this study including a wider range of
countries and cultural groups is needed.
As discussed above, potential cultural influences on beliefs
about sensitive parenting may center around the individualism-
collectivism distinction, and beliefs of parents with individualistic
values converging more with the idea of sensitivity than beliefs
of parents with collectivistic values (Feldman & Masahla, 2010;
Keller, 2013). Further, a horizontal cultural orientation refers to
equality in relations and a vertical orientation refers to social hier-
archies (Singelis, Triandis, Bhawuk, & Gelfand, 1995). Research
has shown that vertical but not horizontal orientations within indi-
vidualism and collectivism were related to authoritarian parenting
(Georgiou, Fousiani, Michaelides, & Stavrinides, 2013). Conversely,
horizontal but not vertical orientations have been found to be associ-
ated with higher autonomy support (Chirkov, Ryan, & Willness,
2005). These findings suggest that horizontal rather than vertical
orientations may be more conducive to sensitive parenting. Another
important aspect of culture is religion. Research regarding religiosity
and parenting to date have yielded inconsistent results and is domi-
nated by Western Christian samples (e.g., Mahoney, Pargament,
Tarakeshwar, & Swank, 2001; Vermeer, 2011). In general, the use
of religion as a guideline in parenting would be expected to poten-
tially lead to less flexible responsiveness, because fixed factors other
than the child’s needs play a role in deciding the response. In the
Emmen et al. (2012) study using the MBQS to assess maternal
beliefs about sensitivity, higher religiosity in childrearing was indeed
related to lower convergence between the mothers’ beliefs about the
ideal mother and the profile of the highly sensitive mother.
There is some evidence that socio-demographic factors are
more salient than cultural factors in predicting sensitive parenting
(Mesman, Van IJzendoorn, & Bakermans-Kranenburg, 2012). Par-
ents with fewer years of formal education and parents with a low
income have been found to have less favorable attitudes about
parenting in general (e.g., Cle´ment & Chamberland, 2009;
Pinderhughes, Bates, Dodge, Pettit, & Zelli, 2000), sensitivity in
particular (Emmen et al., 2012), and show a lower quality of actual
parenting behaviors (e.g., Mesman et al., 2012). Parents experien-
cing socioeconomic hardship may be less inclined to hold views
of parenting that require substantial emotional and time investment
(such as sensitive parenting), because they are more focused on
problems of survival. This may lead these parents to view and
experience parenting more as the stressful day-to-day managing
of children, which is then more likely to foster parenting beliefs
that emphasize the importance of (physical) control rather than
beliefs focusing on the importance of warmth and responsiveness
(McLoyd, 1998; Pinderhughes et al., 2000). Thus, we would expect
higher income and more formal education to relate to maternal
beliefs about parenting that converge with the idea of sensitivity.
2 International Journal of Behavioral Development
at Universiteit Leiden \ LUMC on November 21, 2016jbd.sagepub.comDownloaded from
Other demographic variables of interest include maternal age, fam-
ily size, and urban versus rural residence, with evidence to date
pointing towards less favorable parenting attitudes and behaviors
in younger mothers (e.g., Schlomer & Belsky, 2013), families with
a higher number of children (Furman & Lanthier, 2002), and fam-
ilies in rural areas (Bornstein et al., 2012).
In the current study, we investigate the cross-cultural applicabil-
ity of the sensitivity construct in a sample of 751 mothers from 15
countries with a total of 26 cultural groups. The following hypoth-
eses were tested: (1) There is strong convergence between maternal
descriptions of the ideal mother and attachment theory’s description
of the highly sensitive mother across cultural groups, similar to
findings regarding the overlap between maternal beliefs about the
ideal child and the attachment theory’s notion of a securely attached
child as described by Posada et al. (1995). (2) Convergence between
maternal descriptions of the ideal mother and descriptions of the
highly sensitive mother are predicted by socio-demographic vari-
ables. More specifically, the extent to which mothers describe the
ideal mother as a sensitive mother is expected to show positive asso-
ciations with family income, maternal education, maternal age, and
urbanity of the setting, and a negative association with number of
children. (3) Convergence between maternal descriptions of the ideal
mother and descriptions of the highly sensitive mother are predicted
by cultural variables. More specifically, the extent to which mothers
describe the ideal mother as a sensitive mother is expected to show
positive associations with individualism and a horizontal cultural
orientation, and a negative association with collectivism, a vertical
cultural orientation, and emphasis on religion in parenting.
Method
Sample and procedure
A total of 751 participants were recruited from 15 different coun-
tries representing 26 cultural groups. The countries include (in
alphabetical order): Brazil, Chile, China, Colombia, Egypt, Indone-
sia, Israel, Japan, the Netherlands, Peru, Portugal, Turkey, the
United States, Uruguay, and Zambia. Cultural groups within coun-
tries were defined by the local research teams based on their knowl-
edge of each group’s cultural characteristics. Six countries included
multiple cultural groups (Chile, Israel, the Netherlands, Peru, Por-
tugal, and the US). All participants were selected for having at least
one child between the ages of 6 months and 6 years. Exclusion cri-
teria were a target child with a severe mental or physical disability,
and maternal illiteracy. Recruitment strategies varied across coun-
tries, but generally represented convenience sampling through the
researchers’ networks followed by snowballing, or recruitment of
participants from previous or ongoing other studies. All participants
received the same information brochure (translated into all relevant
languages), and signed the same informed consent form. The bro-
chure informed participants about the international nature of the
study and the research goal of examining differences and similari-
ties about maternal beliefs about parenting between countries.
Small gifts (monetary or otherwise) as a token of appreciation for
participation were adapted to local customs. Socio-demographic
and cultural sample characteristics are shown in Table 1.
Measures
Translation from English into the relevant languages and back-
translation were done by fluent speakers of these languages.
Differences between the English original and the back-translation
were discussed and adaptations to the translation were only made
if the wording was found to be meaningfully different from the
original. For Bahasa Indonesia, no back-translation was done. In
Israel, a dual-focus approach was used in which individuals from
each of the linguistic groups develop the wording of the instructions
and questions simultaneously in order to facilitate equality in clarity
rather than linguistic equivalence (Pen
˜
a, 2007).
Maternal beliefs about sensitive parenting. Maternal views of the
ideal sensitive mother were assessed using the Maternal Behavior
Q-Sort (MBQS; Pederson & Moran, 1995; Pederson, Moran, &
Bento, 1999). The MBQS consists of 90 cards with statements
about maternal behaviors that the mothers sorted into 9 stacks from
‘‘least descriptive’’ (1) to ‘‘most descriptive’’ (9) of the ideal
mother. Because the original items were designed to be evaluated
by professionals rather than mothers, the behavioral descriptions
were simplified for the present study to make them more under-
standable for (low educated) mothers. For example, the item ‘‘Pro-
vides baby with little opportunity to contribute to the interaction’’
was simplified into ‘‘Gives her child little opportunity to play along
or to respond’’ (see Appendix A for the full list of reworded items).
The mothers were first asked to sort the cards into three stacks from
‘‘do not fit the ideal mother at all’’ to ‘‘fit the ideal mother really
well.’’ The mothers were explicitly told that there are no correct
or wrong answers and that it is not about their own parenting beha-
vior, but about what the ideal mother should or should not do. Any
question they had concerning the meaning of an item was answered
according to the item explanations in the protocol. When the moth-
ers distributed the cards across the three stacks, they were asked to
sort each stack into 3 smaller stacks. After the mothers distributed
all cards across nine stacks, they were asked to evenly distribute the
cards across the stacks until each stack consisted of 10 cards.
Consistent with standard Q-sort methodology, each mother’s
sort is represented as an individual variable for data analysis. This
variable consists of 90 cases, representing the 90 cards, with scores
from 1 to 9 reflecting the stack on which the mother put the card. A
mother’s sensitivity belief score was then computed by correlating
her sort with the criterion sort. The criterion sort is provided by the
authors of the MBQS (Pederson et al., 1999), and reflects the highly
sensitive mother. Thus, a higher correlation refers to a greater over-
lap between the mother’s beliefs about the ideal mother and attach-
ment theory’s notion of the highly sensitive mother.
Socio-demographic variables. Maternal educational level was mea-
sured on a 5-point scale: (1) primary school, (2) vocational school,
(3) secondary school/middle vocational education, (4) high voca-
tional education, and (5) and university or higher. Some minor
adjustments were made to this classification system depending on
the local context. Annual gross family income was measured on a
7-point scale that was defined differently in each country based
on the national income distributions. In all countries score
(1) referred to ‘‘no income’’ and score (7) referred to an income
level considered to be very high (and above) in the country of inter-
est. Mothers in Israel were not asked about their income level
because this question would be perceived as violating privacy in the
Israeli context. Mothers also reported on their age and their number
of children. Further, each group was classified as urban versus rural
by the research teams in each country, based on population density
(much lower for rural than urban) and land use (mostly agricultural
or fishing for rural).
Mesman et al. 3
at Universiteit Leiden \ LUMC on November 21, 2016jbd.sagepub.comDownloaded from
Table 1. Socio-demographic and cultural characteristics of the sample.
Family income
(1–7)
Maternal
education
(1–5)
Maternal age
in years
(14–48)
Number of children
(1–6)
Urban (U)/
Rural (R)
Horizontal
Individualism
(1–7)
Horizontal
Collectivism
(1–7)
Vertical
Individualism
(1–7)
Vertical
Collectivism
(1–7)
Religion in
childrearing
(1–5)
Cultural groups (N) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD)
Brazil (15) 3.80 (1.47) 3.33 (1.35) 34.40 (9.25) 1.60 (1.35) U 4.58 (1.48) 5.95 (0.76) 3.02 (1.13) 5.22 (0.96) 2.78 (1.47)
Chile Majority (45) 3.69 (1.47) 3.33 (1.33) 26.80 (6.84) 1.83 (0.97) U 5.39 (1.01) 6.12 (0.70) 3.92 (1.08) 6.36 (0.63) 2.95 (1.29)
Chile-Mapuche (27) 2.22 (0.58) 2.44 (0.70) 28.23 (8.18) 1.96 (1.37) U 5.45 (0.82) 6.07 (0.64) 3.57 (0.98) 6.21 (0.75) 3.44 (1.32)
China (45) 5.05 (1.89) 2.56 (1.27) 31.10 (3.94) 1.40 (0.65) U 5.11 (1.17) 5.39 (0.86) 4.69 (1.07) 6.02 (0.51) –
b
Colombia (40) 2.78 (1.10) 3.30 (1.54) 30.73 (5.73) 1.70 (0.85) U 5.81 (0.78) 6.13 (0.58) 3.83 (0.99) 5.94 (0.88) 3.81 (1.04)
Egypt (12) 3.08 (1.51) 4.25 (0.87) 28.80 (3.74) 2.42 (0.79) U 6.15 (0.95) 6.10 (0.93) 5.33 (1.39) 6.25 (0.72) 4.85 (0.31)
Indonesia (98) 3.42 (1.76) 4.49 (0.65) 32.34 (5.00) 1.64 (0.81) U 5.24 (1.12) 5.37 (0.76) 3.85 (0.85) 6.17 (0.58) 4.56 (0.45)
Israel-Jewish (45) –
a
4.38 (0.49) 33.46 (5.51) 2.00 (0.95) U 5.90 (0.94) 5.82 (0.61) 4.06 (1.30) 5.71 (0.78) 2.68 (1.05)
Israel-Arab (45) –
a
4.84 (0.37) 31.60 (4.58) 2.07 (0.96) U 6.04 (0.67) 6.14 (0.59) 4.46 (1.07) 6.18 (0.71) 3.29 (1.30)
Japan (46) 3.89 (1.49) 3.24 (1.04) 35.82 (5.23) 2.00 (0.94) U 4.49 (0.87) 5.00 (1.06) 3.63 (0.93) 4.98 (0.98) –
b
NL-Dutch (45) 5.51 (1.19) 3.16 (1.18) 33.11 (5.04) 2.11 (0.53) U 5.06 (0.86) 5.69 (0.71) 3.18 (0.96) 4.92 (1.03) 2.18 (1.53)
NL-Turkish (45) 3.92 (1.22) 2.93 (0.88) 29.80 (4.36) 2.13 (0.74) U 5.81 (0.60) 5.79 (0.69) 3.88 (0.85) 6.29 (0.63) 4.54 (0.47)
NL-Moroccan (15) 4.92 (1.22) 3.33 (0.82) 32.20 (4.80) 2.40 (0.63) U 5.60 (0.69) 5.98 (0.59) 3.17 (1.06) 5.60 (0.75) 4.25 (0.53)
NL-Surinamese (15) 4.67 (1.33) 3.60 (0.74) 30.53 (4.02) 1.40 (0.63) U 5.27 (1.12) 5.85 (0.61) 3.06 (0.72) 5.00 (1.20) 2.63 (1.24)
NL-Antillean (15) 3.93 (1.53) 3.00 (0.76) 26.53 (4.45) 1.33 (0.49) U 6.00 (0.92) 5.78 (0.80) 3.35 (0.93) 5.55 (1.06) 3.42 (1.03)
Peru-Majority (15) 6.47 (0.52) 4.99 (0.03) 33.93 (4.18) 1.27 (0.70) U 5.48 (0.79) 6.25 (0.40) 3.85 (1.00) 5.82 (0.36) 3.00 (0.99)
Peru-Lambayeque (15) 2.47 (0.52) 2.60 (0.51) 26.40 (5.60) 2.13 (0.92) R 5.38 (0.73) 5.29 (1.11) 4.42 (1.16) 6.03 (0.95) 3.62 (0.91)
Peru-Puno (15) 5.13 (1.36) 3.73 (1.03) 28.10 (4.71) 1.33 (0.49) R 5.45 (1.27) 4.65 (0.95) 4.68 (1.29) 6.17 (0.91) 3.20 (1.11)
Portugal-Urban (23) 4.39 (1.59) 3.57 (0.84) 32.77 (3.10) 1.38 (0.52) U 4.52 (1.22) 6.03 (0.59) 3.13 (1.28) 6.09 (0.80) 3.24 (1.15)
Portugal-Rural (22) 2.68 (0.57) 2.14 (0.83) 32.09 (4.72) 1.32 (0.48) R 4.90 (1.67) 5.94 (0.63) 3.83 (1.78) 6.37 (0.81) 3.39 (1.05)
Turkey (45) 3.58 (1.70) 2.67 (1.35) 34.10 (5.23) 1.69 (0.70) U 5.53 (1.13) 5.68 (1.05) 4.71 (1.03) 6.12 (0.80) 3.14 (1.19)
US-European (16) 4.94 (1.95) 4.13 (1.02) 36.81 (5.80) 2.69 (0.58) U 5.33 (1.34) 6.08 (0.69) 3.59 (1.38) 5.66 (0.93) 2.45 (1.50)
US-African (15) 4.87 (1.77) 4.33 (0.98) 30.07 (6.65) 1.47 (0.92) U 6.47 (0.52) 5.82 (1.09) 4.00 (1.38) 6.13 (0.71) 2.57 (1.33)
US-Hispanic (15) 4.50 (1.50) 4.00 (1.13) 36.00 (7.04) 2.13 (0.92) U 5.70 (0.97) 6.32 (0.77) 3.83 (1.00) 5.90 (0.94) 2.92 (1.44)
Uruguay (30) 3.80 (1.32) 3.47 (1.68) 32.68 (2.39) 1.57 (0.90) U 5.44 (0.57) 6.05 (0.44) 2.53 (0.96) 5.33 (0.81) 2.03 (1.18)
Zambia (17) 5.12 (1.69) 2.94 (0.90) 29.29 (6.83) 2.18 (1.47) U 4.74 (1.60) 6.13 (0.52) 4.25 (1.39) 6.44 (0.50) 4.44 (0.53)
Note.
a
The measure for family income was not administered in Israel (see method section).
b
The measure for religion in childrearing was not administered in China and Japan (see method section). Family income refers to
annual gross family income on a 7-point scale. In all countries score (1) referred to ‘no income’ and score (7) referred to an income level considered to be very high (and above) in the country of interest. Maternal education
was assessed on a 5-point scale, with small variations between countries: (1) primary school,(2)vocational school, (3) secondary school/middle vocational education,(4)high vocational education, and (5) and university or higher.
Horizontal individualism emphasizes uniqueness and being distinct from the group, vertical individualism emphasizes the distinction in the hierarchy in the form of status, horizontal collectivism focuses on similarities and
common goals with others, and vertical collectivism focuses on sacrificing personal goals for the group. Religion in childrearing refers to the extent to which mothers use their religion in parenting.
4
at Universiteit Leiden \ LUMC on November 21, 2016jbd.sagepub.comDownloaded from
Cultural orientation. Mothers’ cultural orientation was assessed
with a 16-item short version of the Cultural Value Scale (Singelis
et al., 1995; Triandis & Gelfand, 1998), which is a questionnaire
reflecting the dimensions collectivism versus individualism, and
horizontal versus vertical relations. These dimensions form four
scales, each emphasizing a specific cultural orientation (four items
each): horizontal individualism emphasizes uniqueness and being
distinct from the group, vertical individualism emphasizes the dis-
tinction in the hierarchy in the form of status, horizontal collecti-
vism focuses on similarities and common goals with others, and
vertical collectivism focuses on sacrificing personal goals for the
group. The items were rated on a 7-point Likert scale ranging from
strongly disagree (1) to strongly agree (7). Scale scores reflect item
averages. Considering the limited number of items, the internal reli-
abilities of the four subscales were satisfactory (Cronbach’s alphas
.68 to .70). Descriptive statistics of the four cultural orientation
scales are presented in Table 1.
Religion in parenting. Mothers also filled in a questionnaire (devel-
oped for this study) on the extent to which they use their religion in
parenting. This questionnaire consisted of four items rated on a
5-point Likert scale ranging from (1) totally disagree to (5) totally
agree. The items were ‘‘I use my religion as a guideline for the par-
enting of my child,’’ ‘‘My religion helps me to raise my child well,’’
‘‘I teach my child a lot about my religion,’’ and ‘‘I teach my child
that religion plays an important role in our lives.’’ An average item
score was computed for the analyses. When mothers indicated not
having a religion, the final score was set at zero. This questionnaire
was not included in China and Japan, as the questions were deemed
to be inappropriate to the cultural contexts of these countries. The
internal consistency of the scale was high (Cronbach’s alpha ¼
.95). Descriptive statistics of the religion in parenting scale are pre-
sented in Table 1.
Data analysis
Data inspection and analyses to test the hypotheses of the study
were conducted with IMB SPSS Statistics, version 21.0 for
Windows (SPSS Inc., Chicago, IL). Data inspection was performed
within each of the 26 cultural groups. Outliers (z > |3.29|, p ¼ .001,
cf. Tabachnick & Fidell, 2007) were found for MBQ sensitivity
belief score (in total, 6 cases from 5 groups), family size (7 cases
from 5 groups), horizontal individualism (2 cases from 2 groups),
horizontal collectivism (2 cases from 2 groups), vertical collecti-
vism (2 cases from 2 groups), and religion in parenting (1 case).
These values were winsorized to bring them closer to the rest of the
distribution within the relevant groups (Tabachnick & Fidell,
2007).
Missing values were present on most of the predictor variables,
including income (14% missing values), number of children (< 1%
missing values), maternal age (6% missing values), the four cultural
orientation scales (1% missing values on each), and religion in
parenting (13% missing values). Little’s (1988) MCAR (missing
completely at random) test revealed that the assumption of missing
completely at random (Little & Rubin, 2002) was violated,
2
(111)
¼ 338.84, p < 0.01. For variables with more than 5% of cases with
missing values, t tests were conducted to examine whether partici-
pants with missing values differed from those without missing val-
ues on other relevant variables. Income was missing in 14% of the
cases, which was almost entirely due to the fact that income data
was not collected in the two Israel samples (Jewish and Arab),
which accounts for 90 out of 107 missing cases on this variable.
Results of the t tests comparing cases with missing values on
income to other cases therefore represent differences between the
Israel samples and all other samples that can also be derived from
Table 1 (i.e., higher education, more children, higher horizontal
individualism and collectivism). Religion in childrearing was miss-
ing in 13% of cases, which was almost entirely due to the fact that
these questions were not asked in China and Japan (see Methods
section), which accounts for 91 of 99 missing cases on this variable.
Results of the t tests comparing cases with missing values on reli-
gion in childrearing to other cases therefore represent differences
between the Japanese and Chinese samples and all other samples
that can also be derived from Table 1 (i.e., higher maternal age,
lower education due to stratified sampling, lower horizontal indivi-
dualism and collectivism). Finally, maternal age was missing in 6%
of the cases (the majority of which were from Indonesia, Uruguay,
rural Peru and China), and t tests revealed that participants with
missing values on age had significantly fewer children than other
participants.
Because of the violation of MCAR assumption, data were sub-
stituted using multiple imputation under a multilevel model (Rubin,
1987; Van Buuren, 2011). Whereas listwise deletion will only give
unbiased results in statistical analysis when the data are missing
completely at random (MCAR; Little & Rubin, 2002), multiple
imputation will also give unbiased results under the less strict
assumption of missing at random (MAR; Little & Rubin, 2002;
Rubin, 1976). Although it cannot be tested whether the MAR
assumption holds (Schafer, 1997, p. 22), multiple imputation will
still give less biased results than listwise deletion (Schafer, 1997,
p. 26) when this assumption is violated. It was therefore decided
to use multiple imputation for handling the missing data rather than
listwise deletion. Multiple imputation was carried out using the pro-
cedure mice.impute.2 l.norm in R (Van Buuren & Groothuis-
Oudshoorn, 2011). In total, 100 imputed data sets were generated,
and the multilevel regression coefficients and their standard errors
were pooled in SPSS, using Rubin’s combination rules (1987).
2 log likelihoods and likelihood ratio tests were averaged across
the multiply imputed data sets as a rough indication for the specific
model’s fit.
The first hypothesis regarding high convergence between
mothers’ view of the ideal mother and attachment theory’s
notion of sensitive parenting was tested in the f ollowing analy-
ses. First, we computed correla tion coeffic ients for eac h moth-
er’s MBQS sort and the criterion sort that reflects the highly
sensitive parent. In other words: the rankings of the 90 items
by the 751 mothers are used as cases (rather than variables),
so they can be correlated with the criterion sort. High correla-
tions of the mothers’ sorts with the criterion sort would reflect
high convergence between mothers’ notion of the ideal mother
and attachment theory’s notion of the highly sensitive mother.
Second, to further qualify high convergence, we then compared
the average and range of correlations between maternal beliefs
about the ideal mother and attachment theory’s notion o f the
highly sensitive mother (MBQS findings in the current study)
to the average and range found by Posada et al. (1995) with
respect to materna l belie fs about the ideal child and attachment
theory’s notion of the securely attached child, which was
reported to be high. We computed the 84% confidence intervals
(Goldstein & Healy, 1995; Julious, 2004) for our MBQS find-
ings and for Posada et al.’s AQS findings. If the 84% confidence
Mesman et al. 5
at Universiteit Leiden \ LUMC on November 21, 2016jbd.sagepub.comDownloaded from
