See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/51670616

Perceived parenting and risk for major depression in Chinese women

Article in Psychological Medicine · September 2011

DOI: 10.1017/S0033291711001942 · Source: PubMed

CITATIONS READS

39 666

38 authors, including:

Juanhong Gao Yong Li

Beijing University of Chemical Technology Dalian University of Technology
112 PUBLICATIONS 6,767 CITATIONS 819 PUBLICATIONS 16,768 CITATIONS

SEE PROFILE SEE PROFILE

Yuanli Cai Jingkui Chen

Xiangtan University Polytechnique Montréal
101 PUBLICATIONS 3,723 CITATIONS 686 PUBLICATIONS 12,704 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Zhengyi Guo on 30 April 2016.

The user has requested enhancement of the downloaded file.

Psychological Medicine (2012), 42, 921–930. f Cambridge University Press 2011 O R I G I N A L AR T I C LE
doi:10.1017/S0033291711001942

Perceived parenting and risk for major depression

in Chinese women

J. Gao1, Y. Li2, Y. Cai3, J. Chen4, Y. Shen5, S. Ni6, Y. Wei7, Y. Qiu8, X. Zhu1, Y. Liu9, C. Lu10, C. Chen11,
Q. Niu12, C. Tang13, Y. Yang14, Q. Wang15, W. Cui16, J. Xia17, T. Liu18, J. Zhang19, B. Zhao20, Z. Guo21,
J. Pan22, H. Chen23, Y. Luo24, L. Sun25, X. Xiao26, Q. Chen27, X. Zhao28, F. He29, L. Lv30, L. Guo31, L. Liu32,
H. Li33, S. Shi34, J. Flint2, K. S. Kendler34* and M. Tao8*
1
Zhejiang Traditional Chinese Medical Hospital, Hangzhou Zhejiang, P.R. China ; 2 Wellcome Trust Centre for Human Genetics, Oxford, UK ;
3
Fudan University affiliated Huashan Hospital, Shanghai, P.R. China ; 4 Shanghai Jiao Tong University School of Medicine affiliated Shanghai
Mental Health Centre, Shanghai, P.R. China ; 5 Shanghai Tongji University affiliated Tongji Hospital, Shanghai, P.R. China ; 6 Nanjing Brain
Hospital, Nanjing, Jiangsu, P.R. China ; 7 No. 4 Affiliated Hospital of Jiangsu University, Zhenjiang, Jiangsu, P.R. China ; 8 Second Affiliated
Hospital of Zhejiang Chinese Medical University, Hangzhou, Zhe Jiang, P.R. China ; 9 Tianjin Anding Hospital, Hexi District, Tianjin, P.R.
China ; 10 Shandong Mental Health Center, Jinan, Shandong, P.R. China ; 11 No. 1 Hospital of Medical College of Xian Jiaotong University, Xi’an,
Shaanxi, P.R. China ; 12 No. 1 Hospital of Zhengzhou University, Zhengzhou, Henan, P.R. China ; 13 No. 1 Mental Health Center Affiliated Harbin
Medical University, Nangang District, Harbin, Heilongjiang, P.R. China ; 14 Mental Health Center of West China Hospital of Sichuan University,
Wu Hou District, Chengdu, Sichuan, P.R. China ; 15 Beijing Anding Hospital, Capital Medical University, Xicheng District, Beijing, P.R. China ;
16
Hebei Mental Health Center, Baoding, Hebei, P.R. China ; 17 Shengjing Hospital of China Medical University, Heping District, Shenyang,
Liaoning, P.R. China ; 18 Shenzhen Kangning Hospital, Luo Hu, Shenzhen, P.R. China ; 19 No. 3 Affiliated Hospital of Sun Yat-sen University,
Tian He District, Guangzhou, Guangdong, P.R. China ; 20 No. 1 Hospital of Shanxi Medical University, Taiyuan, Shanxi, P.R. China ; 21 Mental
Hospital of Jiangxi Province, Nanchang, Jiangxi, P.R. China ; 22 The First Affiliated Hospital of Jinan University, Tian He District, Guangzhou,
P.R. China ; 23 Wuhan Mental Health Center, Wuhan, P.R. China ; 24 No. 3 Hospital of Heilongjiang Province, Beian, Heilongjiang, P.R. China ;
25
Jilin Brain Hospital, Siping, Jilin, P.R. China ; 26 The First Hospital of China Medical University, He Ping District, Shenyang, P.R. China ;
27
Dalian No. 7 People’s Hospital and Dalian Mental Health Center, Gan Jing Zi District, Dalian, P.R. China ; 28 The First Hospital of Hebei
Medical University, Shijiazhuang, P.R. China ; 29 Lanzhou University Second Hospital, Second Clinical Medical College of Lanzhou University,
Lanzhou, Gansu Province, P.R. China ; 30 Psychiatric Hospital of Henan Province, Xinxiang, Henan, P.R. China ; 31 The Fourth Military Medical
University affiliated Xijing Hospital, Xi’an, Shaanxi, P.R. China ; 32 No. 4 People’s Hospital of Liaocheng, Liaocheng, Shandong, P.R. China ;
33
Guangzhou Brain Hospital/Guangzhou Psychiatric Hospital, Fang Cun Da Dao, Li Wan District, Guangzhou, P.R. China ; 34 Virginia
Commonwealth University, Department of Psychiatry, Virginia Institute for Psychiatric and Behavioral Genetics, Richmond, VA, USA

Background. In Western countries, a history of major depression (MD) is associated with reports of received
parenting that is low in warmth and caring and high in control and authoritarianism. Does a similar pattern exist in
women in China ?

Method. Received parenting was assessed by a shortened version of the Parental Bonding Instrument (PBI) in two
groups of Han Chinese women : 1970 clinically ascertained cases with recurrent MD and 2597 matched controls. MD
was assessed at personal interview.

Results. Factor analysis of the PBI revealed three factors for both mothers and fathers : warmth, protectiveness, and
authoritarianism. Lower warmth and protectiveness and higher authoritarianism from both mother and father were
significantly associated with risk for recurrent MD. Parental warmth was positively correlated with parental
protectiveness and negatively correlated with parental authoritarianism. When examined together, paternal warmth
was more strongly associated with lowered risk for MD than maternal warmth. Furthermore, paternal protectiveness
was negatively and maternal protectiveness positively associated with risk for MD.

Conclusions. Although the structure of received parenting is very similar in China and Western countries, the
association with MD is not. High parental protectiveness is generally pathogenic in Western countries but protective
in China, especially when received from the father. Our results suggest that cultural factors impact on patterns of
parenting and their association with MD.

Received 28 April 2011 ; Revised 24 August 2011 ; Accepted 30 August 2011 ; First published online 27 September 2011

Key words : Major depression, parent–child relations, psychometrics, risk factors, social behaviour.

* Author for correspondence : M. Tao, M.D., Second Affiliated Hospital of Zhejiang Chinese Medical University, No. 318 Chao Wang Road,
Hangzhou, Zhejiang 310005, P.R. China.
(Email : [email protected]) [M. Tao]
(Email : [email protected]) [K. S. Kendler]
The online version of this article is published within an Open Access environment subject to the conditions of the Creative Commons
Attribution-NonCommercial-ShareAlike licence <http://creativecommons.org/licenses/by-nc-sa/2.5/>. The written permission of
Cambridge University Press must be obtained for commercial re-use.
922 J. Gao et al.

Introduction Method
Study subjects
A substantial body of literature suggests a relationship
between the quality of parenting received as a child The data for the present study were drawn from
and risk for psychiatric illness in general and for the ongoing China, Oxford and VCU Experimental
major depression (MD) in particular (Parker, 1979, Research on Genetic Epidemiology (CONVERGE)
1983 ; Burbach & Borduin, 1986 ; Perris et al. 1986, 1994 ; study of MD. CONVERGE studies only women be-
Gerlsma et al. 1990 ; Duggan et al. 1998 ; Kendler et al. cause of evidence that the genetic risk factors for MD
2000). However, almost all of the available studies are not entirely the same in men and women (Kendler
have been conducted in Western countries. Little is et al. 2006). Furthermore, because of evidence that
known of the impact of received parenting on the risk recurrence is the best index of a high familial/genetic
for MD in China. loading for MD (Sullivan et al. 2000), cases were re-
We are aware of three studies that have examined quired to have a history of at least two MD episodes.
the association between experienced parenting and The results presented here are based on a total of
depression in China, only one of which assessed 1970 cases recruited from 51 provincial mental health
the syndrome of MD. Chen (1991) examined 1441 centers and psychiatric departments of general medi-
Taiwanese college students and reported that the cal hospitals in 40 cities in 21 provinces, and 2597
vulnerability to MD increased when family members controls recruited from patients undergoing minor
failed to provide adequate support, when mothers did surgical procedures at general hospitals or from local
not provide consistent love or when fathers were too community centers. All cases and controls were fe-
critical. Kao et al. (1998) explored the relationship male and had four Han Chinese grandparents. Cases
between parenting and current symptomatology in and controls were matched in each collecting center.
1434 adolescent students in Taiwan : strict parenting, Participants provided their Chinese national identity
characterized by high levels of control, was corre- (ID) number, which we used to assign origin to each
lated with depressive symptoms. Finally, Liu (2003) subject. The ID number reflects the place a participant
examined 454 children from Eastern Taiwan. The first applied for an ID card, normally the same as their
Parental Bonding Instrument (PBI ; Parker et al. 1979) household registration (‘ Hu Kou ’) and the place
was used to assess received parenting. The author where they grew up. The majority (68.1 %) of our
failed to replicate the original two-factor structure sample lived in cities, and there was no significant
proposed by Parker et al. (1979) and instead proposed difference in rates of urban versus rural residence
a four-factor solution, with factors termed ‘ caring ’, between cases and controls (67.9 % of cases had an
‘ indifference ’, ‘ autonomy ’ and ‘ overprotection ’. urban address and 68.3 % of controls).
Lower levels of parental care and higher levels of Cases and controls were excluded if they had a pre-
parental indifference were associated with elevated existing history of bipolar disorder, any type of psy-
self-report depressive symptoms. chosis or mental retardation. Cases were aged between
Our aim was to investigate ways in which 30 and 60 years, had had at least two episodes of MD,
received parenting impacts on the liability to MD in with the first episode occurring between age 14 and
mainland China. To do so, we used a cohort of 50 years, and had not abused drug or alcohol before
1970 Chinese women diagnosed with recurrent the first episode of MD. Controls were chosen to match
MD. We first explored whether the dimensions of the region of origin of cases, were aged between
parenting in China, as revealed through the factor 40 and 60 years, had never experienced an episode of
structure of the PBI, resemble those found in MD and were not blood relatives of cases. An older
Western countries. Second, we sought to clarify how minimal age of controls was used to reduce the
these parenting dimensions, taken one at a time chances that they might have a subsequent first onset
and separately in mothers and fathers, impacted of MD. The mean age (and standard deviation) of
on risk for MD. Third, we examined how these in- cases and controls in the dataset was 45.1 (8.8) and 47.7
dividual parenting dimensions altered risk for MD (5.5) years respectively.
when viewed together with the other parenting All subjects were interviewed using a computerized
dimensions within the same parent or the same assessment system that lasted on average 2 h for a case
parenting factor across spouses. Fourth, we were and 1 h for a control. All interviewers were psy-
interested in determining whether the relationship chiatrists or psychiatric nurses and were trained by the
between these factors, taken singly and in combi- CONVERGE team for a minimum of 1 week in the use
nation, and risk for MD was similar in our Chinese of the interview. The interview included assessment of
sample to that seen previously in European and US psychopathology, demographic and personal charac-
populations. teristics, and psychosocial functioning. Interviews
 Perceived parenting and risk for major depression in Chinese women 923

were tape-recorded and a proportion were listened to audio recording of the entire interview, was uploaded
by the trained editors who provided feedback on the to a designated server currently maintained in Beijing
quality of the interviews. by a service provider. All the uploaded files in
The study protocol was approved centrally by the the Beijing server were then transferred to an Oxford
Ethical Review Board of Oxford University and the server quarterly.
ethics committees in participating hospitals in China.
Statistical analysis

Measures Statistical analysis was carried out using the statistical

programming language R (R Development Core
The diagnoses of lifetime MD was established with the
Team, 2004). Before analysis, PBI scores from six items
Composite International Diagnostic Interview [CIDI ;
were reversed for consistency with Parker’s original
World Health Organization (WHO) lifetime version
scoring (Parker et al. 1979). These items are 2, 4, 10, 12,
2.1 ; Chinese version], which classifies diagnoses ac-
14 and 16. A scree plot of a principal components
cording to DSM-IV criteria (APA, 1994). The interview
analysis of the PBI was used to identify factors for
was originally translated into Mandarin by a team of
subsequent analysis. For three dimensions, we picked
psychiatrists in Shanghai Mental Health Center, with
high loading items and summed their responses.
the translation reviewed and modified by members of
All parenting dimensions were centered around their
the CONVERGE team.
means for ease of interpretation. Logistic regression
Parent–child relationships were measured with the
was used to examine the relationship between MD
16-item Parental Bonding Instrument (PBI) modified
and PBI dimensions. Odds ratios (ORs) were obtained
by Kendler (1996) based on the original 25-item in-
by exponentializing the logistic regression coefficients.
strument described by Parker et al. (1979). Parker
(1981) has shown that depressive state does not influ-
ence scores on the PBI. All of the seven original care Results
items from the PBI are included in the 16-item version
Issues in data collection
(items 1, 4, 5, 11, 12, 17 and 18 as originally numbered)
in addition to nine items from the original over- Many interviewers reported difficulties in obtaining
protection scale (items 7, 8, 9, 13, 15, 19, 21, 23 and 25). answers to the PBI from selected respondents.
Information on postnatal depression was assessed Missingness for the PBI data (402/4165 subjects=
using an adaptation of the Edinburgh Scale (Cox et al. 9.7 %) was substantially higher than that observed
1987). with other variables in our project. It was not related to
Education was treated in two ways : as the number general problems of comprehension as missingness
of years of completed full-time education and as an was modestly, but significantly, positively associated
ordinal variable divided into seven levels : no edu- with educational status (p=0.00006) [OR 1.12, 95 %
cation or pre-school (scored 0), primary school or be- confidence interval (CI) 1.06–1.19]. Typically, re-
low (scored 1), junior middle school (scored 2), senior spondents did not so much refuse to report on their
middle school or Technical and vocational school parenting experiences as state that the questions did
(in China, they both take the same time to be finished) not ‘ make sense ’ in the context of their childhood
(scored 3), adult/radio/television schooling, evening experiences. Anecdotes from individual interviewers
education or junior college (scored 4), bachelor degree suggested that, in some of these cases, their inability
(scored 5), master degree or above (scored 6). Social to be able to respond to the PBI items resulted from
class was coded as one of five : (i) executive, business childhood experiences of substantial deprivation
owner or professional ; (ii) administrative personnel, (where parental concerns may have been more focused
clerical or sales worker ; (iii) skilled manual employee ; on basic needs such as providing food than on the
(iv) semi-skilled or unskilled ; (v) other. emotional aspects of parents assessed by the PBI, in
Both the case and control interviews were fully addition to instances of parental abuse and neglect).
computerized into a bilingual system of Mandarin and Although our interview did not assess such depri-
English developed in-house in Oxford, and called vation directly, the finding that missing data on
SysQ. Skip patterns were built into SysQ. Interviews the PBI scales was strongly associated with our
were administered by trained interviewers and en- two measures of childhood adversities supports the
tered offline in real time onto SysQ, which was in- plausibility of this explanation. Among those report-
stalled on laptops. Once an interview was completed, ing childhood neglect, 25.3 % had missing PBI data
a backup file containing all the previously entered versus 7.2 % of those without neglect (x2=25.5, df=1,
interview data could be generated with database p=0.0000004). Among those reporting childhood
compatible format. The backup file, together with an physical abuse, 24.8 % had missing PBI data versus
924 J. Gao et al.

Table 1. Factor loadings for the Parental Bonding Instrument (PBI) in China compared to a US study

China
Kendler (1996)
W P A
W P A
Item number and content F F F M F M F M F

B7 Enjoyed talking things over with me 0.68 x0.06 x0.10 0.60 0.60 0.23 0.29 x0.34 x0.32
B8 Frequently smiled at me 0.61 0.08 x0.15 0.60 0.65 0.29 0.26 x0.37 x0.32
B1 Spoke to me with a warm and friendly voice 0.71 0.08 x0.20 0.55 0.53 0.17 0.14 x0.32 x0.30
B11 Could make me feel better when I was upset 0.68 0.03 x0.17 0.54 0.53 0.33 0.33 x0.37 x0.38
B2 Seemed emotionally cold to mea 0.59 x0.28 0.19 0.41 0.39 x0.10 x0.11 x0.09 x0.13
B3 Appeared to understand my problems and worries 0.69 0.02 x0.14 0.48 0.46 0.23 0.25 x0.34 x0.35
B12 Did not talk with me very mucha 0.54 x0.29 0.14 0.34 0.31 x0.17 x0.14 0.04 0.10
B15 Was overprotective of me 0.09 0.68 0.18 0.21 0.25 0.60 0.55 x0.13 x0.14
B5 Did not want me to grow up x0.08 0.67 x0.01 x0.05 x0.02 0.44 0.44 x0.07 x0.11
B13 Tried to make me dependent on him/her x0.12 0.72 0.01 0.02 0.05 0.57 0.60 x0.05 x0.11
B9 Tended to baby me 0.16 0.73 0.08 0.22 0.28 0.59 0.54 x0.17 x0.14
B6 Tried to control everything I did x0.19 0.60 0.18 x0.24 x0.19 0.41 0.45 0.19 0.13
B16 Let me dress in any way I pleaseda x0.01 0.00 0.67 x0.18 x0.14 x0.14 x0.14 0.49 0.51
B14 Gave me as much freedom as I wanteda 0.00 0.05 0.70 x0.19 x0.17 x0.05 x0.06 0.62 0.64
B10 Let me decide things for myselfa x0.22 0.19 0.67 x0.10 x0.12 x0.05 x0.09 0.80 0.76
B4 Liked me to make my own decisionsa x0.27 0.18 0.58 x0.19 x0.21 x0.01 x0.04 0.71 0.68

W, Warmth ; P, Protectiveness ; A, Authoritarianism ; F, females (mothers of subjects) ; M, males (fathers of subjects).

Numbers in bold in each column represent the loadings used to define the three factors (W, P and A).
a
Loading is reversed on these items for ease of comparability. ‘ Item number ’ is Parker’s original numbering.

7.8 % of those without neglect (x2=12.8, df=1, were significantly more authoritarian [fathers’ mean :
p=0.0003). Finally, significantly more cases failed to 8.3, mothers’ mean : 8.2 (t=2.40, df=4089, p=0.01)],
answer the PBI than controls (13 % v. 5 % respectively ; less protective [fathers’ mean : 8.7 ; mothers’ mean : 9.1
x2=108.0, df=1, p value <2.2r10x16). (t=10.20, df=4088, p<0.00001)] and less warm than
mothers [fathers’ mean : 21.0, mothers’ mean : 21.8
Factor analysis (t=11.99, df=4088, p<0.0001)].

We examined the factor structure of the 16-item PBI, Association of parenting with demographic data
analyzing scores for fathers and mothers separately.
We investigated whether any of the demographic
In both parents, the scree plot suggested three factors
features we had collected on our population were re-
that together explained 39 % and 40 % of the variance
lated to the PBI measures, including urban residence,
in mothers and fathers respectively. Table 1 shows
number of years of occupation and social class.
the loadings on each item and also provides, for
In particular, given the potential importance of the
comparison, the results for fathers from the prior
period 1966–1976 (Cultural Revolution) on families
US sample of Kendler et al. (1997). Factor loadings
(Kleinman, 1986), we looked for an effect of the age of
were reassuringly similar across the two samples.
the subjects on perceived parenting scores. We also
Therefore, we applied the same names to the factors in
wanted to see whether the number of children in
our sample : Warmth, Protectiveness and Authoritar-
the subjects’ families might contribute to variation
ianism. As suggested by the items, high Warmth
in scores, and whether the introduction of China’s
scores reflect a caring and loving parenting style, high
one-child policy might have had an impact on scores.
Protectiveness scores index an overprotective and
Applying a correction threshold for multiple test-
controlling parental style, and high Authoritarianism
ing, we observed no effect of origin (rural versus urban
scores reflect parenting that discourages a child’s
address) on any of the PBI measures. Similarly, we
sense of independence and autonomy.
found no significant effect of years of education, nor of
social status, on any PBI measure. We also found no
Maternal versus paternal parenting
effect of age on any PBI measure, nor any significant
Mothers and fathers differed in their Warmth, differences in PBI measures for those who would have
Protectiveness and Authoritarianism scores. Fathers been children (under the age of 18) during the Cultural
 Perceived parenting and risk for major depression in Chinese women 925

Table 2. Odds ratios (ORs) for the association between paternal and maternal parenting dimensions and major depression (MD)
examined separately, both one at a time and all together

Analysed separately Analysed together

Parent PBI factor p OR 95 % CI p OR 95 % CI

Father Warmth 1.5r10x46 0.62 0.58–0.66 1.5r10x28 0.64 0.59–0.69

Authoritarianism 4.8r10x23 1.38 1.30–1.48 0.0006 1.14 1.05–1.22
Protectiveness 2.3r10x7 0.84 0.79–0.90 0.70 1.01 0.94–1.09
Mother Warmth 6.8r10x36 0.67 0.62–0.71 8.4r10x20 0.70 0.65–0.76
Authoritarianism 8.4r10x22 1.37 1.28–1.45 3.3r10x5 1.17 1.08–1.26
Protectiveness 0.02 0.93 0.87–0.99 0.28 1.03 0.96–1.12

ORs with associated 95 % confidence intervals (CIs) and p values (p) for the effect of three Parental Bonding Instrument (PBI)
dimensions on MD. Factors were standardized before analysis so that the OR represents the increase in risk for MD associated
with an increase of one standard deviation in the relevant parenting dimension.

Revolution, compared to those who were either born but remained highly significant in both parents. By
afterwards or were older than 18 at that period. Six contrast, the impact of Protectiveness on risk for MD
percent of our sample were singletons, but we found disappeared in both mother and father.
no effect of the number of children in the subject’s
family on the reported PBI measures. Paternal and maternal effects considered jointly
The ratings of received parenting from mothers
Association of parenting with risk for MD and fathers were highly correlated : Warmth +0.67,
Authoritarianism +0.77 and Protectiveness +0.84
Main effects and interactions
(all highly significant at p<0.0001). It was therefore of
We began by testing for an association between each interest to examine jointly the impact of paternal and
PBI factor individually and risk for MD separately in maternal Warmth, Authoritarianism and Protective-
mothers and fathers. As shown in Table 2, in both ness. We first examined the main effects and then, in a
parents, high levels of Warmth and Protectiveness re- separate analysis, explored the significance of any
duced and high levels of Authoritarianism increased possible interactions.
susceptibility for MD. The protective effect of Warmth As shown in Table 3, when considered jointly,
was slightly greater in fathers than in mothers and father’s Warmth had a considerably stronger protec-
a similar but somewhat larger difference was seen tive impact on risk for MD (OR 0.69) than did mother’s
for Protectiveness. The pathogenic effects of high Warmth (OR 0.82). Of note, a modestly significant
Authoritarianism were nearly identical in the two positive interaction was also observed between pa-
parents. ternal and maternal warmth. Interpreted in a positive
The three dimensions of parenting were signifi- direction, this means that, on the logistic scale, the
cantly correlated. In fathers, the correlations were as protective effect on risk for MD of having both a
follows (all highly significant at p<0.00001) : Warmth– mother and father with high Warmth was modestly
Protectiveness +0.36 Warmth–Authoritarianism greater than that predicted from the sum of the indi-
x0.50 and Protectiveness–Authoritarianism x0.23. vidual effects in mother and father taken on their own.
In mothers, the correlations were broadly similar When considered jointly, the pathogenic effect of
(all highly significant at p<0.00001) : Warmth– maternal Authoritarianism was slightly stronger than
Protectiveness +0.29, Warmth–Authoritarianism that of paternal Authoritarianism and there was no
x0.51 and Protectiveness–Authoritarianism x0.15. evidence of an interaction. For Protectiveness, the re-
It was therefore of interest to determine, using sults were very different. When examining paternal
multiple regression analyses, the effect of each di- and maternal effects jointly, paternal Protectiveness
mension of parenting controlling for the other two remained protective, with an effect size (0.73) mod-
dimensions. These results are also shown in Table 2. estly greater than that observed when examined on its
In these joint analyses, the protective effect of own (0.84). However, maternal Protectiveness chan-
Warmth on risk for MD was only slightly attenuated ged sign and moved from being modestly protective
in both fathers and mothers. For Authoritarianism, when examined on its own (OR 0.92) to being associ-
the association with risk for MD declined substantially ated with an increased risk for MD when paternal
926 J. Gao et al.

Table 3. Odds ratios (ORs) for the association between paternal and maternal parenting dimensions and major depression (MD)
examined together, first estimating only main effects and then estimating the interaction

Dimension Parameter Father Mother Interaction

Warmth OR 0.69 0.82 0.95

95 % CI 0.63–0.75 0.75–0.90 0.89–0.99
p 8.13r10x17 8.99r10x6 0.048
Authoritarianism OR 1.19 1.23 0.99
95 % CI 1.07–1.31 1.11–1.36 0.94–1.03
p 0.0007 7.58r10x5 0.53
Protectiveness OR 0.73 1.19 1.02
95 % CI 0.64–0.83 1.05–1.36 0.98–1.06
p 1.09r10x6 0.006 0.22

ORs with associated 95 % confidence intervals (CIs) and p values (p) for the effect of three Parental Bonding Instrument (PBI)
factors on MD. Factors were standardized before analysis so that the OR represents the increase in risk for MD associated with
an increase of one standard deviation in the relevant parenting dimension. Two models were run. The first examined only the
main effects and those estimates are presented in the table. The second examined the main effects and the interaction and the
estimates of the interaction from that analysis are depicted in the table.

Protectiveness was included in the model (OR 1.19). We began by examining the factor structure of the
Again, no significant interaction was observed. PBI, the instrument we used to assess parenting. We
identified in our Chinese sample a three-factor sol-
Interactions between parenting dimensions ution with individual factors representing parental
Warmth, Protectiveness and Authoritarianism. This
Finally, we examined interactions across dimensions, three-factor PBI solution found by Kendler et al. (1997)
with a particular interest in evaluating Parker’s hy- has been replicated in two other US samples (Cox et al.
pothesis of the pathogenic effect for depression of 2000 ; Lizardi & Klein 2002), and also in Japan (Sato
what he termed ‘ affectionless control ’ : the combi- et al. 1999) and Brazil (Terra et al. 2009). Although
nation of low parental Warmth and high parental based on only a single sample and a single instrument,
Protectiveness (Parker, 1979). Of the six possible two- our results are consistent with the hypothesis that
way interactions between the three parenting dimen- the broad dimensions of parenting are similar across
sions, examined separately in mothers and fathers, culturally diverse human populations.
one was significant (p<0.001) and consistent across We assessed the independent association between
both parents, a positive interaction between Author- the three parenting factors, separately in mothers and
itarianism and Protectiveness. That is, in the presence fathers, and the risk for MD. Consistent with a wide
of high Authoritarianism (which is risk predisposing) array of prior studies (Parker, 1979 ; Perris et al. 1986 ;
and high Protectiveness (which is protective), the risk Plantes et al. 1988 ; Parker & Hadzi-Pavlovic, 1992 ;
for MD is higher than would have been predicted by Oakley-Browne et al. 1995 ; Rey, 1995 ; Rodgers,
the main effects of the two parenting dimensions con- 1996a, b ; Kendler et al. 2000), the most robust associ-
sidered one at a time. Of the four remaining possible ation we observed was an inverse one between
interactions, only one was significant and that at a Warmth and depressive risk. The magnitude of the
marginal level (p=0.03). In fathers only, a positive association can be perhaps be best expressed by re-
interaction was observed between Warmth and versing the dimension and calling it parental coldness
Authoritarianism such that the joint effect of the two (Kendler et al. 1997). In this case, a one standard devi-
produced a higher risk for MD than would be pre- ation increase in coldness from mother or father was
dicted from the two dimensions taken one at a time. associated with a 50–60 % increase in risk for MD,
modestly greater than the 38–47 % increased risk ob-
served in a US female twin population that included
Discussion
exactly this same scale analyzed in the same manner
We examined, in a large sample of Chinese women, (Kendler et al. 2000).
the relationship between risk for recurrent MD and Similar to other studies in Western populations
three dimensions of retrospectively reported parent- (Oakley-Browne et al. 1995 ; Rey 1995 ; Rodgers
ing received from their mothers and their fathers. 1996a, b ; Kendler et al. 2000), we also found that high
 Perceived parenting and risk for major depression in Chinese women 927

levels of Authoritarianism were associated with an The three levels of parenting were highly correlated
increased risk for MD. The ORs seen in our sample in mothers and fathers, most strongly for Protective-
(1.38 in fathers and 1.37 in mothers) were very similar ness and least for Warmth. It was therefore of interest
to those seen in the US twin sample studied by to examine the impact of each parent’s rearing prac-
Kendler et al. (2000) (1.33 in fathers and 1.34 in tices, taking account of that of their spouse. These
mothers). analyses could also examine interactions between
However, in our sample of Chinese women, paternal and maternal parenting styles, some of which
high levels of Protectiveness from both mothers and have been hypothesized to be particularly pathogenic.
fathers, when analyzed independently, were associ- A parenting style characterized by high levels of
ated with a decreased risk for MD (ORs of 0.84 and control and protectiveness and low levels of warmth
0.93 respectively). This is the opposite pattern of that (referred to as ‘ affectionless control ’) has been hy-
observed in Western samples. For example, in the pothesized to be particularly pathogenic with respect
Virginia twins, high Protectiveness was associated to risk for MD (Parker 1979 ; Mackinnon et al. 1993 ;
with an increased risk of MD, significantly in mothers Sato et al. 1998). In addition, a warm loving relation-
(OR 1.26) and non-significantly in fathers (OR 1.14). Of ship with one parent might mitigate the impact of a
note, for all three of our dimensions, the association poor relationship with the other parent. If so, this
with MD was slightly stronger for paternal than for would predict an interaction between maternal and
maternal parenting, suggesting that for women in paternal Warmth in the prediction of MD.
China the father has a psychological role of at least When parental and maternal parenting dimensions
as much importance in the child’s life as does the were examined jointly, we found that the association
mother. between paternal Warmth and MD was stronger than
We next observed correlations in the dimensions of that found for maternal Warmth. Most surprisingly,
received parenting separately in mothers and fathers. when examined jointly, high paternal Protectiveness
The positive correlations between Protectiveness remained protective and associated with a reduced
and Authoritarianism (+0.36 in fathers and +0.29 in risk for MD. However, when controlling for levels
mothers) were similar to the averaged correlation re- of paternal Protectiveness, high maternal Protective-
ported by the Virginia sample (+0.31). However, in ness, now more consistent with results in Western
our sample the inverse correlation between Warmth countries, becomes associated with an elevated risk
and Authoritarianism was stronger (about x0.50) of MD.
than that observed in the Virginia twin (x0.33). Of The examination of interactions between parenting
particular interest, whereas in the Virginia sample was generally less informative. As seen previously
Warmth and Protectiveness were modestly negatively in the Virginia twin sample (Kendler et al. 2000), we
correlated (x0.11), in our Chinese sample these two found no evidence for the concept of ‘ affectionless
dimensions of parenting were positively correlated control ’ that the combination of low parental Warmth
(+0.36 in fathers and +0.29 in mothers). These results and high parental Protectiveness or Authoritarianism
together suggest an important cultural difference in was more pathogenic than the main effects of these
parenting between the USA and China. In the USA, parenting dimensions examining in isolation.
parental Protectiveness is inversely related to Warmth Our findings can be usefully compared with those
and increases risk for MD in offspring, whereas in of the one other study in East Asia examining the as-
China, parental Protectiveness is positively related to sociation between PBI and lifetime MD of which we
Warmth and decreases risk for MD. are aware. In 418 employed Japanese adults, Narita
We also examined the joint effect of all three et al. (2000) found support for the three-factor solution
parenting dimensions on risk for MD. Although the of the PBI that we also observed in this study. The
impact of Warmth was slightly attenuated, the effects strongest association they observed was between low
of Authoritarianism were substantially reduced and parental Warmth and risk for MD. Controlling for
Protectiveness completely eliminated. These results levels of Warmth, no significant effects were seen for
are similar to those observed in the Virginia female Authoritarianism (which they had labeled in reverse
twin sample (Kendler et al. 2000). In both samples, direction ‘ encouragement of behavioral freedom ’).
the most robust and consistent relationship between One significant effect was seen for Protectiveness
parenting and risk for MD was parental Warmth. (which they had labeled ‘ denial of psychological
However, as noted earlier, whereas in China high autonomy ’). In opposition to our findings, they found
parental Protectiveness can be thought of as a proxy that high paternal Protectiveness increased risk for
for parental warmth and lovingness, in the USA high MD in women. Our findings, which require confir-
parental Protectiveness is a proxy for low levels of mation, suggest interesting differences in Chinese
parental warmth and care. women that may not be shared with other East Asian
928 J. Gao et al.

populations, in the meaning of Protectiveness when co-workers have explored the relationship between
received from their father. suicide in depression in China, demonstrating the
There are several ways that perceived parenting importance of cultural factors in contributing to psy-
could be linked to susceptibility to MD. For example, chiatric disease (Phillips et al. 2002 ; Zhang et al. 2004,
parenting abilities of mothers with postnatal de- 2010). Our work shows how factors known to predis-
pression could be perceived differently from those pose to MD in the West operate in China, sometimes in
who do not suffer postnatal depression. We were not expected ways, sometimes unexpected.
able to explore whether perinatal depression in the
mothers of cases could be a link between onset of MD
Limitations
and perceived maternal parenting, but we were able to
explore the impact of perinatal depression in cases These results should be interpreted in the context of
themselves, and found no detectable effect (Tian et al. four potentially important methodological limitations.
2011). First, the sample is entirely female and our results may
To what extent do our findings reflect differences in or may not extrapolate to men in China. Second, we
parenting styles between Asian and Western parents ? were missing PBI data on a non-trivial proportion of
Although there is a dearth of studies relevant to adult our sample, with much higher missingness in our
psychopathology in China, there has been interest in cases than our controls. Anecdotal evidence suggests
the effects of parenting on child development and that many of these cases had childhoods that were
school achievement, providing instructive comparison so deprived and stressful that the traditional PBI
with our results. Typically, these investigations use questions about parent–child relationships were not
a different parenting typology from ours, distinguish- meaningful. The constructs of parenting assessed by
ing between authoritarian and authoritative styles the PBI were not sensible for them in the conditions in
(Baumrind, 1971). The former is characterized by low which they grew up. These missing data, however,
warmth, restricting the child’s independence, and the are more likely to lead to an underestimation of the
frequent use of coercive discipline, thus conceptually association between poor parenting on risk for MD
similar to parents scoring low on the Warmth and high than an overestimation.
on the Authoritarianism scale of the PBI ; the latter Third, because of the retrospective nature of our
is characterized by high warmth and acceptance, re- data, we could not determine the degree to which the
specting the child’s autonomy and setting reasonable association between reporting parenting and history
limits on behavior. These features are a combination of for MD is causal or a result of (a) retrospective
high scores on the Warmth and Protectiveness scales recall bias, (b) passive gene–environment correlation
derived from the PBI. in which depressed individuals parent poorly and
Most studies examining the effect of parenting on pass on their risk genes to their children or (c) other
children’s academic, social and psychological adjust- confounding factors that might predispose to both
ment have found patterns similar to those obtained in poor parenting and MD. Correlations between con-
the West (Chen et al. 1997, 2000 ; Lai & McBride-Chang, temporaneously recorded child-rearing practices and
2001), but some have argued that Western categoriza- later recall are moderate (McCrae & Costa, 1988 ; Dunn
tions of parenting styles do not capture the full range & McGuire, 1994), as are the levels of agreement be-
of parenting variation found in mainland China tween relatives in retrospectively reported parenting
(Chao, 1994, 2000 ; Stewart et al. 1998 ; Wu et al. 2002). (Parker, 1981, 1983 ; Schwarz et al. 1985). Reports
Our findings for the effect of Protectiveness as a of retrospective parenting are generally stable over
proxy for parental warmth and lovingness, rather than time (Parker, 1989 ; Wilhelm et al. 2005). Although
increasing risk for MD as it does in the West, may re- skepticism about retrospective reporting of parenting
flect a mixed authoritarian and authoritative parenting is certainly appropriate, we suggest that the reports
style. on parenting analyzed here probably have sufficient
Our results add to the discussion about the simila- reliability and validity to be seriously considered as a
rities and dissimilarities between MD in China and reflection of true parenting and a true causal influence
the rest of the world. Kleinman has pointed out that on risk for MD (McCrae & Costa, 1988 ; Parker, 1989).
‘ culturally coded symptoms may confound diagnosis ’ Fourth, in this report we have emphasized com-
(Kleinman, 2004) and has written extensively about parisons between data from our current sample and
the difference in the ways mood is conceptualized and those obtained in the Virginia twins. This is a useful
expressed in China (Kleinman, 1986). Lee (1999) points comparison because both samples were entirely fe-
out how psychiatric diagnoses, including MD, are in male, and the studies used the same PBI items and
part the product of vested interests, political strategies closely related statistical methods. However, it is im-
and social policies in China ; Phillips, Zhang and portant to note that, whereas the Virginia sample is
 Perceived parenting and risk for major depression in Chinese women 929

population based, the Chinese sample is clinically as- Dunn J, McGuire S (1994). Young children’s nonshared
certained and the cases of MD in our Chinese sample experiences : a summary of studies in Cambridge
would be expected on average to be more severe. and Colorado. In Separate Social Worlds of Siblings :
The Impact of Nonshared Environment on Development
(ed. E. M. Hetherington, D. Reiss and R. Plomin),
Acknowledgements pp. 111–128. Lawrence Erlbaum Associates : Hillsdale, NJ.
Gerlsma C, Emmelkamp PMG, Arrindell WA (1990).
This work was funded by the Wellcome Trust. The Anxiety, depression, and perception of early parenting :
authors are part of the CONVERGE consortium and a meta-analysis. Clinical Psychology Review 10, 251–277.
gratefully acknowledge the support of all partners in Kao YM, Wu CI, Lue BH (1998). The relationships between
hospitals across China. inept parenting and adolescent depression dimension
and conduct behaviors. Chinese Journal of Family Medicine 8,
11–21.
Declaration of Interest Kendler KS (1996). Parenting : a genetic-epidemiologic
perspective. American Journal of Psychiatry 153, 11–20.
None.
Kendler KS, Gatz M, Gardner CO, Pedersen NL (2006).
A Swedish national twin study of lifetime major
depression. American Journal of Psychiatry 163, 109–114.
References
Kendler KS, Myers J, Prescott CA (2000). Parenting and
APA (1994). Diagnostic and Statistical Manual of Mental adult mood, anxiety and substance use disorders in
Disorders. American Psychiatric Association : Washington, female twins : an epidemiological, multi-informant,
DC. retrospective study. Psychological Medicine 30, 281–294.
Baumrind D (1971). Current patterns of parental authority. Kendler KS, Sham PC, MacLean CJ (1997). The
Developmental Psychology Monograph 4, 1–103. determinants of parenting : an epidemiological,
Burbach DJ, Borduin CM (1986). Parent-child relations and multi-informant, retrospective study. Psychological
the etiology of depression : a review of methods and Medicine 27, 549–563.
findings. Clinical Psychology Reviews 6, 133–153. Kleinman A (1986). Social Origins of Distress and Disease.
Chao RK (1994). Beyond parental control and authoritarian Depression, Neurasthenia and Pain in Modern China. Yale
parenting style : understanding Chinese parenting University Press : New Haven, CT.
through the cultural notion of training. Child Development Kleinman A (2004). Culture and depression. New England
65, 1111–1120. Journal of Medicine 351, 951–953.
Chao RK (2000). The parenting of immigrant Chinese and Lai KW, McBride-Chang C (2001). Suicidal ideation,
European American mothers : relations between parenting parenting style, and family climate among Hong Kong
styles, socialization goals, and parental practices. Journal of adolescents. International Journal of Psychology 36, 81–87.
Applied Developmental Psychology 21, 233–248. Lee S (1999). Diagnosis postponed : shenjing shuairuo and
Chen RC (1991). Vulnerability of factors of depression among the transformation of psychiatry in post-Mao China.
college students in Taiwan. Chinese Journal of Mental Health Culture, Medicine and Psychiatry 23, 349–380.
5, 53–69. Liu Y-L (2003). Parent–child interaction and children’s
Chen X, Dong Q, Zhou H (1997). Authoritative and depression : the relationships between parent–child
authoritarian parenting practices and social and school interaction and children’s depressive symptoms in
performance in Chinese children. International Journal of Taiwan. Journal of Adolescence 26, 447–457.
Behavioral Development 21, 855–873. Lizardi H, Klein DN (2002). Evidence of increased sensitivity
Chen X, Liu M, Li B, Cen G, Chen H, Wang L (2000). using a three-factor version of the Parental Bonding
Maternal authoritative and authoritarian attitudes and Instrument. Journal of Nervous and Mental Disease 190,
mother–child interactions and relationships in urban 619–623.
China. International Journal of Behavioral Development 24, Mackinnon A, Henderson AS, Andrews G (1993). Parental
119–126. ‘ affectionless control ’ as an antecedent to adult
Cox BJ, Enns MW, Clara IP (2000). The Parental Bonding depression : a risk factor refined. Psychological Medicine 23,
Instrument : confirmatory evidence for a three-factor 135–141.
model in a psychiatric clinical sample and in the National McCrae RR, Costa PT (1988). Do parental influences
Comorbidity Survey. Social Psychiatry and Psychiatric matter ? A reply to Halverson. Journal of Personality 56,
Epidemiology 35, 353–357. 445–449.
Cox JL, Holden JM, Sagovsky R (1987). Detection of Narita T, Sato T, Hirano S, Gota M, Sakado K, Uehara T
postnatal depression. Development of the 10-item (2000). Parental child-rearing behavior as measured by the
Edinburgh Postnatal Depression Scale. British Journal of Parental Bonding Instrument in a Japanese population :
Psychiatry 150, 85–100. factor structure and relationship to a lifetime history of
Duggan C, Sham P, Minne C, Lee A, Murray R (1998). depression. Journal of Affective Disorders 57, 229–234.
Quality of parenting and vulnerability to depression : Oakley-Browne MA, Joyce PR, Wells JE, Bushnell JA,
results from a family study. Psychological Medicine 28, Hornblow AR (1995). Adverse parenting and other
185–191. childhood experience as risk factors for depression in
 930 J. Gao et al.

women aged 18–44 years. Journal of Affective Disorders 34, Parental Bonding Instrument in a Japanese population.
13–23. Psychological Medicine 29, 127–133.
Parker G (1979). Parental characteristics in relation to Sato T, Sakado K, Uehara T, Narita T, Hirano S, Nishioka K,
depressive disorders. British Journal of Psychiatry 134, Kasahara Y (1998). Dysfunctional parenting as a risk factor
138–147. to lifetime depression in a sample of employed Japanese
Parker G (1981). Parental reports of depressives. An adults : evidence for the ‘ affectionless control ’ hypothesis.
investigation of several explanations. Journal of Affective Psychological Medicine 28, 737–742.
Disorders 3, 131–140. Schwarz JC, Barton-Henry ML, Pruzinsky T (1985).
Parker G (1983). Parental Overprotection : A Risk Factor in Assessing child-rearing behaviors : a comparison of ratings
Psychosocial Development. Grune & Stratton : New York. made by mother, father, child, and sibling on the CRPBI.
Parker G (1989). The Parental Bonding Instrument : Child Development 56, 462–479.
psychometric properties reviewed. Psychiatric Stewart SM, Rao N, Bond MH (1998). Chinese dimensions of
Developments 7, 317–335. parenting : broadening Western predictors and outcomes.
Parker G, Hadzi-Pavlovic D (1992). Parental representations International Journal of Psychology 33, 345–358.
of melancholic and non-melancholic depressives : Sullivan PF, Neale MC, Kendler KS (2000). Genetic
examining for specificity to depressive type and for epidemiology of major depression : review and
evidence of additive effects. Psychological Medicine 22, meta-analysis. American Journal of Psychiatry 157,
657–665. 1552–1562.
Parker G, Tupling H, Brown L (1979). A Parental Bonding Terra L, Hauck S, Schestatsky S, Fillipon AP, Sanchez P,
Instrument. British Journal of Medical Psychology 52, 1–10. Hirakata V, Ceitlin LH (2009). Confirmatory factor
Perris C, Arrindell WA, Eisemann ME (1994). Parenting and analysis of the Parental Bonding Instrument in a Brazilian
Psychopathology. John Wiley & Sons : New York. female population. Australian and New Zealand Journal of
Perris C, Arrindell WA, Perris H, Eisemann M, van der Psychiatry 43, 348–354.
Ende J, von Knorring L (1986). Perceived depriving Tian T, Li Y, Xie D, Shen Y, Ren J, Wu W, Guan C, Zhang Z,
parental rearing and depression. British Journal of Zhang D, Gao C, Zhang X, Wu J, Deng H, Wang G, Zhang
Psychiatry 148, 170–175. Y, Shao Y, Rong H, Gan Z, Sun Y, Hu B, Pan J, Sun S,
Phillips MR, Yang G, Zhang Y, Wang L, Ji H, Zhou M Song L, Fan X, Zhao X, Yang B, Lv L, Chen Y, Wang X,
(2002). Risk factors for suicide in China : a national Ning Y, Shi S, Kendler KS, Flint J, Tian H (2011). Clinical
case-control psychological autopsy study. Lancet 360, features and risk factors for post-partum depression in a
1728–1736. large cohort of Chinese women with recurrent major
Plantes MM, Prusoff BA, Brennan J, Parker G (1988). depressive disorder. Journal of Affective Disorders. Published
Parental representations of depressed outpatients from a online 6 August 2011. doi:10.1016/j.jad.2011.06.047.
U.S.A. sample. Journal of Affective Disorders 15, 149–155. Wilhelm K, Niven H, Parker G, Hadzi-Pavlovic D (2005).
R Development Core Team (2004). A Language and The stability of the Parental Bonding Instrument over a
Environment for Statistical Computing. R Foundation for 20-year period. Psychological Medicine 35, 387–393.
Statistical Computing : Vienna. Wu P, Robinson CC, Yang C, Hart CH, Olsen SF, Porter CL,
Rey JM (1995). Perceptions of poor maternal care are Jin S, Wo J, Wu J (2002). Similarities and differences in
associated with adolescent depression. Journal of Affective mothers’ parenting of preschoolers in China and the
Disorders 34, 95–100. United States. International Journal of Behavioral Development
Rodgers B (1996a). Reported parental behaviour and adult 26, 481–491.
affective symptoms. 1. Associations and moderating Zhang J, Conwell Y, Zhou L, Jiang C (2004). Culture,
factors. Psychological Medicine 26, 51–61. risk factors and suicide in rural China : a psychological
Rodgers B (1996b). Reported parental behaviour and adult autopsy case control study. Acta Psychiatrica Scandinavica
affective symptoms. 2. Mediating factors. Psychological 110, 430–437.
Medicine 26, 63–77. Zhang J, Li N, Tu XM, Xiao S, Jia C (2010). Risk factors for
Sato T, Narita T, Hirano S, Kusunoki K, Sakado K, rural young suicide in China : a case-control study. Journal
Uehara T (1999). Confirmatory factor analysis of the of Affective Disorders 129, 244–251.

View publication stats