r Human Brain Mapping 37:3757–3766 (2016) r

Pathway to Neural Resilience: Self-Esteem

Buffers Against Deleterious Effects of Poverty
on the Hippocampus

Yinan Wang,1 Lin Zhang,2 Xiangzhen Kong,3 Yingyi Hong,4

Bobby Cheon,5,6 and Jia Liu1*
1
School of Psychology, Beijing Key Laboratory of Applied Experimental Psychology, Beijing
Normal University, Beijing, China
2
Civil Aviation Medical Center, Civil Aviation Administration of China, Beijing, China
3
State Key Laboratory of Cognitive Neuroscience and Learning & IDG/McGovern Institute for
Brain Research, Beijing Normal University, Beijing, China
4
Business School, The Chinese University of Hong Kong, Hong Kong, China
5
Division of Psychology, Nanyang Technological University, Singapore
6
Clinical Nutrition Research Centre, Singapore Institute for Clinical Sciences (A*STAR),
Singapore

r r

Abstract: Human neuroimaging studies have shown that people living in poverty tend to suffer hippo-
campal atrophy, which leads to impaired memory and learning throughout life. However, behavioral
studies demonstrate that poor people with high self-esteem are often exempt from the deleterious
effect of poverty and instead possess a happy and successful life. Here we investigated whether high
self-esteem can buffer against the deleterious effects of poverty, as indicated by low subjective socioe-
conomic status (SSS), on the hippocampal gray matter volume (GMV) in a large cohort of young par-
ticipants (N 5 280). As expected, findings revealed that although low (vs. high) SSS was linked with a
smaller hippocampal GMV, the deleterious effect of low SSS on hippocampal GMV was alleviated
when the participants have high self-esteem. Commonality analyses further confirmed this observation.
The current study suggests that positive psychological resources such as self-esteem may provide
protection for the hippocampal atrophy in adversity. Hum Brain Mapp 37:3757–3766, 2016. C 2016
V
Wiley Periodicals, Inc.

Key words: socioeconomic status; self-esteem; hippocampus; resilience

r r

The first two authors contributed equally to this work.
Contract grant sponsor: National Natural Science Foundation of
China; Contract grant number: 31230031; Contract grant sponsor:
National Basic Research Program of China; Contract grant num-
ber: 2014CB846101; Contract grant sponsor: National Natural Sci-
ence Foundation of China; Contract grant numbers: 31221003,
31471067, 31470055; Contract grant sponsor: National Social
Science Foundation of China; Contract grant number:
13&ZD073,14ZDB160; Contract grant sponsor: Changjiang Schol-
ars Programme of China.
*Correspondence to: Jia Liu, Room 1415, School of Psychology, 19
Xinjiekouwai St, Haidian District, Beijing 100875, China. E-mail:
liujia@bnu.edu.cn
Received for publication 19 September 2015; Revised 16 May
2016; Accepted 17 May 2016.
DOI: 10.1002/hbm.23273
Published online 30 May 2016 in Wiley Online Library
(wileyonlinelibrary.com).

C 2016 Wiley Periodicals, Inc.

V
 r Wang et al.
INTRODUCTION
In 2012, nearly one billion children, or every second
child in the world, lived in poverty (cf. The Human Devel-
opment Report, 2012), suffering various stressful life
events, such as violence, hunger, poor health and abuse
[Baum et al., 1999]. It has long been recognized that pov-
erty has a deleterious effect not only on physical health
but also on mental health and cognitive abilities [Adler
et al., 1994; Conger and Donnellan, 2007]. Indeed, prior
studies have revealed that the hippocampus, a critical neu-
ral substrate for regulating responses under stress, is
adversely affected by chronic stress [Avishai-Eliner et al.,
2001; Fenoglio et al., 2006; Gianaros et al., 2007; Plotsky
and Meaney, 1993]. Given that poverty and low socioeco-
nomic status (SES) may be chronic sources of stress and
adversity [Baum et al., 1999], it has also been implicated
as a risk factor for problems with hippocampal structure
and function. Structural magnetic resonance imaging
(MRI) studies have revealed smaller hippocampus in chil-
dren from low-income families [Hanson et al., 2011; Jed-
norog et al., 2012], and such association remains detectable
even more than 50 years later [Staff et al., 2012]. Higher
SES, as reflected by educational attainment, may also serve
as a buffer against age-related reductions in hippocampal
volume [Noble et al., 2012]). Because accumulating evi-
dence from meta-analyses converges to demonstrate the
pivotal role of the hippocampus in memory and learning
[K€uhn and Gallinat, 2014; Martinelli et al., 2013], atrophy
of the hippocampus may consequently further prevent
children in low-income families from higher achievement
that could help them escape the cycle of poverty.
Previous studies have demonstrated that psychosocial
resources can greatly buffer the deleterious effect of pov-
erty [Creswell et al., 2005]. One such psychological
resource is self-esteem, which refers to one’s overall sub-
jective emotional evaluation of his or her own worth [Pan
et al., 2016; Rosenberg, 1965]. That is, self-esteem can
either be positive (i.e., high self-esteem) or negative (i.e.,
low self-esteem). In the Kauai longitudinal Study,
researchers tracked the development of 698 children born
in 1955 on the island of Kauai in Hawaii. They found that
children who lived in poverty but had high levels of self-
esteem developed into stable, competent, confident, and
productive adults [Werner, 1995]. A similar result was
Abbreviations
FWHM Full width at half maximum
GLM General linear model
GM Gray matter
GMV Gray matter volume
IQR Interquartiles
MP-RAGE Magnetization prepared rapid gradient echo
MRI Magnetic resonance imaging
SES Socioeconomic status
SSS Subjective socioeconomic status
r 3758
r
observed in successful youth who grew up in London’s
high-crime neighborhoods in the Longitudinal Cambridge
Study [Werner, 1997]. Interestingly, individuals with high
self-esteem tend to have larger hippocampal volumes
[Kubarych et al., 2012; Pruessner et al., 2005]. Given the
protective role of self-esteem against poverty in previous
studies, we hypothesized that high self-esteem promotes
hippocampal resilience, especially in individuals in
poverty.
To test this hypothesis, a large population of college stu-
dents was recruited (N 5 280), and voxel-based morphom-
etry (VBM) was used to examine whether the
hippocampus of low SES participants was modulated by
self-esteem. Specifically, we first replicated the relationship
between participants’ hippocampal gray matter volume
(GMV) and their social economic status, such that a higher
SES would be linked with larger hippocampal GMV. Sec-
ond, we measured participants’ self-esteem with the
Rosenberg Self-Esteem Scale [Rosenberg, 1965] and tested
our prediction that high self-esteem can buffer the deleteri-
ous effects of poverty on hippocampal GMV.
MATERIALS AND METHODS
Participants
Two hundred and eighty college students (138 males;
mean age 5 22.68 years, SD 5 1.07 years) from Beijing Nor-
mal University, China, participated in this study as a part
of an ongoing project investigating relation among genes,
environment, brain and behavior [Kong et al., 2016; Song
et al., 2014; Wang et al., in press]. Data that are irrelevant to
the scope of this study were not reported in this study. The
participants reported no past or current psychiatric illness
or history of neurological disorders (e.g., epilepsy, traumatic
brain injury, neurodegenerative disorders and cerebro-
vascular disease), mental retardation or significant systemic
medical illness. Both the behavioral and MRI protocols
were approved by the Institutional Review Board of Beijing
Normal University. Written informed consents were
obtained from all participants prior to the experiment.
Behavioral Measures
Socioeconomic Status (SES)
Participants’ SES was assessed by the MacArthur Scale
(youth version) of Subjective Social Status [Goodman
et al., 2001], which measures individuals’ subjective per-
ception of their families’ place on the social ladder. The
subjective socioeconomic status (SSS) was selected because
perceived social standing is more closely associated with
stress-related mental and physical health than objective
SES (e.g., education, income and occupation) [Adler et al.,
2000; Singh-Manoux et al., 2005]. In this study, partici-
pants were presented a ten-rung “social ladder” and
r
 r Self-Esteem Promotes Hippocampal Resilience
asked to indicate the point that “best represents where
your family was” during their development (i.e., child-
hood [6–12 years old], adolescent [12–18 years old] and
youth [18–22 years old]) compared with other families in
China. The participants were informed that at the top of
the ladder was the most well-to-do family, with the most
money, education and respected jobs, whereas families at
the bottom were the worst off.
Self-Esteem
Participants’ self-esteem was assessed in Chinese with
an established and validated Chinese version of Rosenberg
Self-Esteem Scale [Wang et al., 2010; Yang and Wang,
2007], which contains ten items to assess one’s global self-
worth, incorporating both positive and negative feelings
about the self (e.g., “On the whole, I am satisfied with
myself”). The participants were instructed to report the
extent to which they agreed or disagreed with each state-
ment using a 6-point Likert-type scale (1 5 strongly dis-
agree, 6 5 strongly agree). The total score of self-esteem
was calculated by summing the scores of each item, and a
higher score indicated a more positive evaluation of self-
worth and value.
MRI Acquisition
Participants were scanned using a Siemens 3T scanner
(MAGENTOM Trio with a Tim system) with a 12-channel
phased-array head coil at the BNU Imaging Center for
Brain Research, Beijing, China. MRI structural images
were acquired using a 3D magnetization prepared rapid
gradient echo (MP-RAGE) T1-weighted sequence (TR/TE/
TI 5 2530/3.39/1100 ms, flip angle 5 7 degrees, FOV 5 256
3 256 mm2). One hundred and twenty-eight contiguous
sagittal slices were imaged with 1 3 1-mm in-plane resolu-
tion and 1.33-mm slab thickness for whole brain coverage.
Image Processing for Voxel-Based Morphometry
VBM was employed to characterize the neuroanatomical
differences in GMV and neuroanatomical correlates of
behavioral performance across participants [Ashburner
and Friston, 2000]. Specifically, VBM was performed using
SPM8 (Statistical Parametric Mapping, Wellcome Depart-
ment of Imaging Neuroscience, London, UK) and Diffeo-
morphic Anatomical Registration through Exponentiated
Lie Algebra (DARTEL, Wellcome Department of Imaging
Neuroscience). First, image quality was assessed by visual
examination. Second, the origin of each brain was man-
ually set to the anterior commissure for each participant.
Third, the images were segmented into four distinct tissue
classes: gray matter (GM), white matter, cerebrospinal
fluid and everything else (e.g., skull and scalp) using a
unified segmentation approach [Ashburner and Friston,
2005]. Fourth, the GM images for each participant were
r 3759
r
rigidly aligned and resampled to 2 3 2 3 2 mm. Fifth, the
images were nonlinearly registered with DARTEL, which
involves iteratively computing a study-specific template
based on the tissue maps from all participants and then
warping all participants’ GM images into the generated
template to increasingly improve the alignment [Ash-
burner, 2007]. Sixth, the GM images were normalized to
standard MNI space, and the GM voxel values were
modulated by multiplying the Jacobian determinants
derived from the registration to preserve the volume of tis-
sue from each structure [Good et al., 2002]. The modulated
GM images were then smoothed with an 8-mm full width
at half maximum (FWHM) isotropic Gaussian kernel.
Finally, to exclude noisy voxels, the modulated images
were masked using an absolute masking with a threshold
of 0.2. The masked modulated GM images were used for
further statistical analyses.
Exclusion Criteria
In the measurement of self-esteem, outliers were defined
as two interquartiles (IQR) below the first quartile or two
IQR above the third quartile. One participant (male) was
excluded because his self-esteem score was two IQRs below
the first quartile. Another participant (male) was excluded
from further analysis because of missing items in the ques-
tionnaire. In the measurement of neural substrate, another 6
participants (male, 2% of the sample) whose images had
excessive scanner artifacts or showed gross anatomical
abnormalities were excluded. Therefore, among all 280 par-
ticipants tested, 272 participants (male: 130, age mean
5 22.64, SD 5 1.03 years) were include in the VBM analysis.
Statistical Analysis of VBM
Because the GM images were normalized to the stand-
ard MNI space, we defined the whole hippocampus using
probabilistic maps from the Harvard-Oxford Subcortical
Structural Atlas implemented in FSL [the FMRIB Software
Library, www.fmrib.ox.ac.uk/fsl (Smith et al., 2004)], and
only included voxels that have 25% or greater probability
of being labeled as the hippocampus. The total GMV of
the right/left hippocampus was calculated by summariz-
ing the probability values (or GMV) of all voxels within
the right/left hippocampal mask from the modulated
images. We also obtained the total hippocampal GMV by
adding hippocampal GMVs of both hemispheres together.
To examine the deleterious effect of poverty on hippo-
campal GMV, an independent two-sample t-test was con-
ducted to compare hippocampal GMV between the
participants with high SSS (i.e., above the median) and the
participants with low SSS (i.e., below the median). Because
the disputed nature of the definition of poverty, the result
found in the low SSS group was used to infer the deleteri-
ous effect of poverty on the hippocampus. Sex and the
total GMV of the whole brain were treated as confounding
r
 r Wang et al.
Figure 1.
The distribution of SSS and the effect of SSS on hippo-
campal GMV. (A) The histogram shows the distribution of
participants’ SSS. High-SSS (dark gray) and low-SSS (light gray)
were identified as those individuals who scored above or below
the median, respectively. (B) The graph shows the effect of SSS
on the hippocampal GMV. Participants in low-SSS had signifi-
covariates, which were regressed out from the variance of
hippocampal GMV. Similar analyses were also conducted
on hippocampal GMV in the right and left hemispheres
separately to confirm the correlation between SSS and hip-
pocampal GMV. Considering the participants tested were
college students, who had a narrow range of ages (mean
age 5 22.64 years, SD 5 1.03 years); therefore, we did not
regress out the confounding factor of age.
To further examine whether the hippocampus was the
primary and major brain region dampened by poverty, a
whole-brain t-test analysis was performed by comparing
the GMV of each voxel between low SSS and high SSS
group. Sex and the total GMV of the whole brain were
also treated as confounding covariates. Correction for mul-
tiple comparisons was applied using a non-stationary
whole-brain cluster-level correction (min t > 3.29, cluster
significance: P < 0.01) [Hayasaka et al., 2004].
Finally, we explored whether high self-esteem could
buffer against the deleterious effect of poverty on hippo-
campal GMV. To this end, a general linear model (GLM)
was performed to test the effects of SSS, self-esteem, and
their interaction on participants’ hippocampal GMV, while
controlling participants’ sex and their total GMV of the
whole brain. The participants’ self-esteem scores were
mean-centered before entering into the GLM. In addition,
commonality analyses [Nimon, 2010] were used to assess
the predictive contributions of SSS and self-esteem to
explain individual differences in hippocampal GMV in
high and low SSS groups, respectively. In each commonal-
ity analysis, there were two predictor variables: SSS and
self-esteem. Thus there were two unique variance contrib-
utors (USSS, Uself-esteem) and one common variance contrib-
r 3760
r
cantly less hippocampal GMV than the high-SSS group. The y
axis indicates the residual of hippocampal GMV after regressing
out the variance of gender and the total GMV of the whole
brain. Asterisks indicate a significant difference between groups
(**P < 0.001).
utor (CSSS, self-esteem). To calculate the variance
contributions, three regression models were built. In the
three models, SSS, self-esteem, and the combination of SSS
and self-esteem were set as the independent variables,
respectively. Consistently, the dependent variable in all
three models was the residual of hippocampal GMV after
regressing out the variance of sex and the GMV of the
whole brain. The total variance explained in each regres-
sion model was labeled R1, R2, or R12, respectively. Thus,
the unique variance contribution of SSS (USSS) was calcu-
lated by subtracting R2 from R12, while the unique var-
iance contribution of self-esteem (Uself-esteem) was
computed by subtracting R1 from R12. The common var-
iance contribution of SSS and self-esteem (CSSS, self-esteem)
was calculated as (R11 R2 – R12). To evaluate the relative
influence of SSS and self-esteem, the proportion of each
contributor in the total explained variance was calculated.
To further investigate whether the self-esteem interacted
with SSS specifically to predict hippocampal GMV, we
examined if the association between SSS and GMV in
other regions that were smaller for low SSS participants
also interacted with self-esteem.
RESULTS
Relationship Between SSS and
Hippocampal GMV
Because the SSS values in the three periods
(i.e., childhood, adolescent, and youth) were highly corre-
lated (rchildhood & adolescent 5 0.82; radolescent & youth 5 0.87;
rchildhood & youth 5 0.67, all ps < 0.001), the SSS values of the
r
 r Self-Esteem Promotes Hippocampal Resilience
Figure 2.
The whole-brain analysis based on the contrast between
high versus low SSS group. The whole-brain analysis revealed
two clusters in the bilateral hippocampus where the high SSS
group possessed a larger GMV than the low SSS group. Also
shown in the figure is that the high SSS group possessed less
GMV than the low SSS group in the medial prefrontal cortex.
The color bar indicates Z values. [Color figure can be viewed at
wileyonlinelibrary.com]
three periods were averaged to simplify the analysis and
comprehensively represent participants’ SSS during their
course of growth, with a lower score of SSS indicating a
more impoverished upbringing. There SSS scores ranged
from 1 to 8.67 (mean 5 4.21, median 5 4, SD 5 1.64)
(Fig. 1A).
Furthermore, we validated the measurement of SSS
using participants’ objective SES index by their household
wealth. Participants were asked whether their families
possess following items (landline telephone, refrigerator,
washing machine, air conditioner, kitchen, DVD player,
internet, the classics, poetry, artwork, private car, bath-
room, computer, mobile phone, TV) in their childhood,
adolescent and youth. Participants responded either yes or
no, and the yes responses were summed so that a higher
score indicated higher socioeconomic circumstances in that
period. There were significant correlations between partici-
pants’ SSS scores and their objective SES at different peri-
ods (childhood: r 5 0.655, P < 0.001; adolescent: r 5 0.641,
P < 0.001; youth: r 5 0.635, P < 0.001), confirming the par-
ticipants with lower SSS scores did come from more
impoverished families.
r 3761
r
To replicate the previous finding that the hippocampus
was impaired by poverty [Hanson et al., 2011; Jednorog
et al., 2012; Staff et al., 2012], we explored the relationship
between SSS and hippocampal GMV. To balance the num-
ber of participants between high and low SSS groups, par-
ticipants whose SSS scores were equal or below the median
were categorized as the low SSS group (N 5 144, males: 72;
mean age 5 22.89 years; mean SSS 5 2.96, SD 5 0.87),
whereas those with the scores higher than the median were
classified as the high SSS group (N 5 128, males: 50; mean
age 5 22.38 years; mean SSS 5 5.63, SD 5 1.03). The hippo-
campus was defined using probabilistic maps from the
Harvard-Oxford Subcortical Structural Atlas available for
FSL, including only voxels that had 25% or greater proba-
bility of being labeled as the hippocampus. We compared
hippocampal GMV between the participants with high SSS
and the participants with low SSS, with sex and GMV of
the whole brain treated as covariates of confounding fac-
tors, and we found that the low SSS group had significantly
less hippocampal GMV than the high SSS group
[t (270) 5 3.77, P < 0.001, Cohen’ d 5 0.46; Fig. 1B].
To examine whether the pattern was present in the hip-
pocampus of each hemisphere, we performed the same
analysis on each hippocampus respectively. A similar pat-
tern was observed in both the right [t (270) 5 3.31,
P 5 0.001, Cohen’ d 5 0.40] and left [t (270) 5 3.70,
P < 0.001, Cohen’ d 5 0.45] hippocampus. After controlling
for age, as well as sex and whole-brain GMV, the regres-
sion coefficient of hippocampal GMV and SSS remained
significant (b5 0.15, P < 0.001), which was similar to the
results without controlling for age.
In addition, a whole-brain regression analysis where SSS
was treated as a continuous variable also revealed a signif-
icant correlation between SSS and hippocampal GMV
(Z 5 3.29, P 5 0.01; peak coordinates: x 5 24, y 5 214,
z 5 218), similar to the result with the analysis based on
the contrast of high versus low SSS groups. Thus, our
results replicate previous findings that poverty is associ-
ated with smaller GMV of the hippocampus.
To further explore whether the rest of the brain is
affected by SSS, a whole-brain t-test was performed by
TABLE I. Regions that showed significantly differences
in GMV between low SSS and high SSS participants
MNI coordinate
Cluster
Regions size t-max x y z
Right hippocampus 3352 4.87 224 228 26
Left hippocampus 2552 4.65 26 214 216
Superior parietal lobule 992 4.5 16 246 74
Medial prefrontal 20968 26.00 212 56 210
cortex *
Precentral gyrus * 1608 24.57 24 216 50
Note: t > 3.29, cluster significance: P < 0.01; “*”: low SSS group
had a larger GMV in this region.
r
 r Wang et al. r

Figure 3.
Self-esteem was associated with hippocampal GMV only for the low SSS participants, but not for
the high SSS participants.

comparing the GMV of each voxel between low SSS and predicted SSS 3 self-esteem interaction effect was margin-
high SSS group, with sex and the total GMV of the whole ally significant [t (266) 5 21.93, P 5 0.055, gp2 5 0.01]. To
brain as confounding covariates. We found that the high unpack the interaction effect, we calculated the partial cor-
SSS participants had greater GMV in bilateral hippocam- relations between self-esteem and hippocampal GMV after
pus (see Fig. 2 and Table I) and superior parietal lobule controlling for participants’ sex and the total GMV of the
(SPL), and smaller GMV in medial prefrontal cortex and whole brain for the low and high SSS group, respectively.
precentral gyrus than low SSS participants. As predicted, the partial correlation between self-esteem
and hippocampal GMV was significant only for the low
SSS participants (r 5 0.24, P < 0.01, Fig. 3), but not for the
Interaction Effect of SSS and Self-Esteem
high SSS participants (r 5 0.002, P 5 0.99). Meanwhile, the
on Hippocampal GMV
effect of self-esteem on the hippocampus of the low SSS
Previous behavioral studies have shown that high self- group was significant larger than that of the high SSS
esteem provides protection against the deleterious effect of group (Steiger’s Z 5 1.98, P < 0.05, one tail), suggesting
low SES [Werner, 1995, 1997]; here we further investigated that self-esteem especially buffers the deleterious effect of
whether high self-esteem enhances the resilience of the hip- poverty on the hippocampus of individuals with low SSS.
pocampus to poverty. To this end, we conducted a General To further examine the effects of SSS and self-esteem
Liner Model (GLM) to predict participants’ hippocampal had hippocampal GMV, commonality analyses were used
GMV using SSS and self-esteem and their interaction term to assess the predictive contributions of SSS and self-
as predictors, and sex and the total GMV of the whole brain esteem in explaining the variance in hippocampal GMV of
as covariates. Participants’ self-esteem was measured by the
Rosenberg Self-esteem Scale [Rosenberg, 1965]. The coeffi- TABLE II. Results of commonality analyses
cient alpha for the sample was 0.88, showing reasonably
high reliability in assessment of the participants’ self-esteem. Variance Variance Proportion in the total
contributor explained (R2) explained Variance (%)
The kurtosis and skewness of all scores ranged from 21
and 11, indicating the normality of the data [Marcoulides Low SSS group
and Hershberger, 1997]. Meanwhile, the total score for the USSS 0.0040 6.3
ten items on the scale ranged from 29 to 60 (mean 5 44.70, Uself-esteem 0.055 85.9
SD 5 6.45), showing considerable individual differences. CSSS,self-esteem 0.0050 7.8
Results revealed that the main effect of SSS was signifi- Total 0.064 100
cant [t (266) 5 3.36, P 5 0.001, gp2 5 0.04], which is consist- High SSS group
ent with the t-test we reported above. The main effect of USSS 0.0016 89.2
self-esteem was also significant [t (266) 5 2.25, P < 0.05, Uself-esteem 0.000042 2.3
CSSS,self-esteem 0.00015 8.5
gp2 5 0.02], showing that participants with higher self-
Total 0.0018 100
esteem had larger hippocampal GMV. Importantly, the

r 3762 r
 r Self-Esteem Promotes Hippocampal Resilience r

Figure 4.
The contributions of SSS and self-esteem toward explaining the variance of hippocampal
GMV in low SSS and high SSS participants. The pie charts show the proportions of variance
explained by SSS and self-esteem in the total explained variance of hippocampal GMV in low SSS and
high SSS group, respectively. The area indicates the effect from different variance contributors.

participants from high and low SSS groups. The results of
the commonality analyses are shown in Table II. In the
low SSS group, the variance observed in hippocampal
GMV was best explained by the unique variance of self-
esteem (R2 5 0.055, 85.9% of total variance explained by
the SSS and self-esteem together), followed by the shared
variance of SSS and self-esteem (R2 5 0.005, 7.8%) and the
unique variance of SSS (R2 5 0.004, 6.3%). In contrast, the
variance of hippocampal GMV of the participants from
high SSS families was best explained by the unique var-
iance of SSS (R2 5 0.0016, 89.2%, Fig. 4), followed by the
shared contribution of SSS and self-esteem (R2 5 0.00015,
8.5%) and the unique variance of self-esteem
(R2 5 0.000042, 2.3%). That suggests that material resources
from families (i.e., SSS) play a dominant role in shaping
hippocampal GMV when individuals live in an advan-
taged environment, whereas hippocampal GMV is largely
modulated by psychological resources (i.e., self-esteem) if
individuals are in an improvised environment.
Finally, to investigate whether self-esteem interacted with
SSS specifically to predict hippocampal GMV, we examined
whether the association between SSS and GMV in the SPL,
another region associated with SSS, also interacted with
self-esteem. We only observed a significant main effect of
SSS on SPL GMV [t (266) 5 4.30, P < 0.001], not the main
effect of self-esteem [t (266) 5 20.32, P 5 0.75]. Meanwhile,
the interaction effect of SSS by self-esteem in the hippocam-
pus was significantly different from that in the SPL
(Steiger’s Z 5 21.98, P < 0.05), suggesting that the interac-
tion was specific to hippocampal GMV.
DISCUSSION
The present study investigated the role of self-esteem in
the resilience of the hippocampus to the deleterious effects
of poverty during development in a large cohort of young
adult participants. As expected, participants with rela-
tively low SSS had smaller hippocampal GMV than those
with relatively high SSS. More importantly, self-esteem
played a buffering role, such that the effect of SSS on hip-
pocampal GMV depended on the participant’s level of
self-esteem. In the low SSS group, participants with high
self-esteem showed larger hippocampal GMV than did
their counterparts with similar SSS but low self-esteem. In
contrast, self-esteem had no effect on hippocampal GMV
of participants from the high SSS group. Further, the var-
iance of hippocampal GMV was mainly explained by self-
esteem among low SSS participants, whereas SSS played a
dominant role among the high SSS participants. In addi-
tion, the finding that self-esteem did not interact with SSS
to predict GMV in other cortical regions is consistent with
this stress-related hypothesis. Therefore, self-esteem likely
provides protection for the hippocampus, specifically for
those growing up in a disadvantaged environment.
The finding that participants’ SSS level was positively
correlated with hippocampal GMV is consistent with pre-
vious studies [Jednorog et al., 2012; Staff et al., 2012]. The
deleterious effect of low SSS likely comes from chronic
stress [Baum et al., 1999], which leads to excessive secre-
tion of stress hormones such as neurotoxic glucocorticoids
[Hackman et al., 2010; McEwen and Gianaros, 2010].
Because the hippocampus contains a high density of glu-
cocorticoids receptors, it appears particularly susceptible
to the deleterious effect of low SSS [Avishai-Eliner et al.,
2001; McEwen, 2012; Plotsky and Meaney, 1993]. Indeed,
prior studies have revealed that the hippocampus, a criti-
cal neural substrate for regulating responses under stress,
is adversely affected by chronic stress [Avishai-Eliner
et al., 2001; Fenoglio et al., 2006; Gianaros et al., 2007; Plot-
sky and Meaney, 1993].

r 3763 r
 r Wang et al.
Importantly, our study demonstrates for the first time
that self-esteem can improve the resilience of the hippocam-
pus to low SSS. The underlying mechanism of this protec-
tion is unclear, but one possibility is that self-esteem may
protect the hippocampus by reducing sensitivity to stress
and thus decreasing the secretion of neurotoxic stress hor-
mones [Blum, 1998; Creswell et al., 2005; Goldenberg et al.,
2001]. Indeed, high self-esteem can lead to ‘positive illu-
sion,’ or positive self-evaluation, which is an exaggerated
perception of control and unrealistic optimism [Taylor and
Brown, 1994]. Previous neuroimaging work shows that self-
esteem may modulate neural responses to social feedbacks
in brain regions such as dorsal anterior cingulate cortex,
dorsal medial prefrontal cortex [Eisenberger et al., 2011],
and the ventral anterior cingulate cortex [Somerville et al.,
2010]. Recently, self-esteem is also found to be related to
brain network properties between the medial prefrontal
cortex and the ventral striatum [Chavez and Heatherton,
2015]. Therefore, self-esteem likely emerges from neural
systems integrating information about the self with positive
reward and self-perception. Such positive perception and
expectation may help lowering susceptibility to perceived
stress and thus reducing physiological responses to stress
[Bonanno et al., 2005; Crocker and Park, 2004; Hostinar
et al., 2014; Mantzicopoulos, 1990]. As a result, the hippo-
campus is less likely to be exposed to excessive levels of
neurotoxic glucocorticoids in individuals because of the
protection of self-esteem [Pruessner et al., 2005]. Impor-
tantly, the finding that self-esteem didn’t interact with SSS
to predict GMV in other regions that were smaller for low
SSS participants strengthened our stress-related hypothesis.
For future study, only with serial longitudinal assessments
of self-esteem and structural brain images along with con-
current reports of perceived stress, we can identify whether
self-esteem provides protection for hippocampus through
modifying the expression of stress-related variations in hip-
pocampal and cortical morphology [Feder et al., 2009; Feno-
glio et al., 2006].
The finding that self-esteem promotes hippocampal resil-
ience to poverty has important practical implications. First,
our study suggests potential solutions to help individuals
suffering poverty. For example, the hippocampus may
maintain normal development if people of low SES can
maintain positive self-regard through exercise, setting real-
istic expectations, or spending time with friends [Mruk,
2006]. In addition to self-esteem, other external protective
factors, such as parenthood relationships, social support,
sense of control, and positive thinking style, have been
found to provide protection for individuals from low SES
families [Chen and Miller, 2013; Hostinar et al., 2014]. It
would be interesting to explore whether these factors also
promote hippocampal resilience to poverty, and, if they do,
how both internal and external factors work in concert to
resist the deleterious effects poverty has on the brain.
Besides the association between SSS and the hippocam-
pus, we also observed a significant negative correlation
r 3764
r
between SSS and the GMV of the mPFC, suggesting that
lower SSS individuals had a larger GMV of the mPFC. This
finding is in principle consistent with previous studies that
have linked childhood SES with adult PFC volume [Cohen
et al., 2006; Gianaros et al., 2007; Lawson et al., 2013; Noble
et al., 2015] and PFC activity in both cognitive [Sheridan
et al., 2012] and emotional tasks [Javanbakht et al., 2015].
This association may be accounted for by the role of the
mPFC in emotional processing, especially emotional regula-
tion [Etkin et al., 2011]. Indeed, chronic stress during child-
hood is found to mediate the relationship between
childhood poverty and the activity of prefrontal cortical
region [Kim et al., 2013; Sheridan et al., 2012]. Therefore, it
is possible that children from low-SES family need to
employ more emotional regulation to handle with the stress
from poverty, which might lead to a larger mPFC volume.
In other words, individuals with lower SSS are likely to
possess larger mPFC volume as we observed here.
In summary, this study demonstrates that self-esteem
protects the hippocampus against the deleterious effects of
poverty, which suggests practical implications for professio-
nal intervention policies. In addition, this study provides a
neurobiological framework for investigating the association
between psychophysiological development and self-esteem
in adverse environments. In our study, several issues
remained unaddressed and are important topics for future
research. First, given the correlational nature of the results
in our study, the cause-and-effect relationship between SES,
self-esteem and hippocampal GMV remains unclear. There-
fore, our stress-related hypothesis is proposed simply based
on previous studies [Hanson et al., 2011; Jednorog et al.,
2012; Noble et al., 2012; Werner, 1995, 1997]. Future longitu-
dinal studies are needed to directly illustrate how self-
esteem modulates the association between SES and hippo-
campal GMV, and to uncover the critical period of psycho-
physiological development during which the effect of self-
esteem is most pronounced [Chen and Miller, 2013; Curtis
and Cicchetti, 2003; Werner, 1995]. One possible experiment
is to compare individuals’ stress levels and hippocampal
volumes before and after intervention aimed at improving
their self-esteem levels so that we can directly test whether
self-esteem protects the hippocampus through stress-related
pathway. Second, we focused on the hippocampus as the
primary region of interest in our study. Since several
regions such as medial prefrontal cortex are also modulated
by SES, further studies are expected to examine the rela-
tionship between the SES, GMV of these regions and other
psychological variables.
REFERENCES
Adler NE, Boyce T, Chesney MA, Cohen S, Folkman S, Kahn RL,
Syme SL (1994): Socioeconomic status and health. The chal-
lenge of the gradient. Amer Psychol 49:15–24.
Adler NE, Epel E, Casterllazzo G, Ickovics J (2000): Relationship
of subjective and objective social status with psychological and
r
 r Self-Esteem Promotes Hippocampal Resilience
physical health in healthy white women. Health Psychol 19:
586–592.
Ashburner J (2007): A fast diffeomorphic image registration algo-
rithm. Neuroimage 38:95–113.
Ashburner J, Friston KJ (2000): Voxel-based morphometry—The
methods. Neuroimage 11:805–821.
Ashburner J, Friston KJ (2005): Unified segmentation. Neuroimage
26:839–851.
Avishai-Eliner S, Eghbal-Ahmadi M, Tabachnik E, Brunson KL,
Baram TZ (2001): Down-regulation of hypothalamic
corticotropin-releasing hormone messenger ribonucleic acid
(mRNA) precedes early-life experience-induced changes in
hippocampal glucocorticoid receptor mRNA. Endocrinology
142:89–97.
Baum A, Garofalo J, YALI A (1999): Socioeconomic status and
chronic stress: Does stress account for SES effects on health?
Ann N Y Acad Sci 896:131–144.
Blum M (1998): Healthy youth development as a model for youth
health promotion: A review. J Adolesc Health 22:368–375.
Bonanno GA, Rennicke C, Dekel S (2005): Self-enhancement
among high-exposure survivors of the September 11th terrorist
attack: Resilience or social maladjustment? J Person Soc Psy-
chol 88:984.
Chavez RS, Heatherton TF (2015): Multimodal frontostriatal con-
nectivity underlies individual differences in self-esteem. Soc
Cogn Affect Neurosci 10:364–370.
Chen E, Miller GE (2013): Socioeconomic status and health: Medi-
ating and moderating factors. Ann Rev Clin Psychol 9:723–749.
Cohen RA, Grieve S, Hoth KF, Paul RH, Sweet L, Tate D,
Gunstad J, Stroud L, McCaffery J, Hitsman B (2006): Early life
stress and morphometry of the adult anterior cingulate cortex
and caudate nuclei. Biol Psychiatry 59:975–982.
Conger RD, Donnellan MB (2007): An interactionist perspective
on the socioeconomic context of human development. Ann
Rev Psychol 58:175–199.
Creswell JD, Welch WT, Taylor SE, Sherman DK, Gruenewald TL,
Mann T (2005): Affirmation of personal values buffers neuro-
endocrine and psychological stress responses. Psychol Sci 16:
846–851.
Crocker J, Park LE (2004): The costly pursuit of self-esteem. Psy-
chol Bull 130:392–414.
Curtis WJ, Cicchetti D (2003): Moving research on resilience into
the 21st century: Theoretical and methodological considera-
tions in examining the biological contributors to resilience.
Dev Psychopathol 15:773–810.
Eisenberger NI, Inagaki TK, Muscatell KA, Haltom KEB, Leary
MR (2011): The neural sociometer: Brain mechanisms underly-
ing state self-esteem. J Cognit Neurosci 23:3448–3455.
Etkin A, Egner T, Kalisch R (2011): Emotional processing in ante-
rior cingulate and medial prefrontal cortex. Trend Cognit Sci
15:85–93.
Feder A, Nestler EJ, Charney DS (2009): Psychobiology and molec-
ular genetics of resilience. Nat Rev Neurosci 10:446–457.
Fenoglio KA, Brunson KL, Baram TZ (2006): Hippocampal neuro-
plasticity induced by early-life stress: Functional and molecu-
lar aspects. Front Neuroendocrinol 27:180–192.
Gianaros PJ, Jennings JR, Sheu LK, Greer PJ, Kuller LH, Matthews
KA (2007): Prospective reports of chronic life stress predict
decreased grey matter volume in the hippocampus. Neuro-
image 35:795–803.
Goldenberg JL, Pyszczynski T, Greenberg J, Solomon S, Kluck B,
Cornwell R (2001): I am not an animal: Mortality salience,
r 3765
r
disgust, and the denial of human creatureliness. J Exp Psychol
Gen 130:427.
Good CD, Johnsrude IS, Ashburner J, Henson RN, Fristen K,
Frackowiak RS (2002). A voxel-based morphometric study of ageing
in 465 normal adult human brains. Paper presented at the Bio-
medical Imaging, 2002. 5th IEEE EMBS International Summer
School on.
Goodman E, Adler NE, Kawachi I, Frazier AL, Huang B, Colditz
GA (2001): Adolescents’ perceptions of social status: Develop-
ment and evaluation of a new indicator. Pediatrics 108:E31.
Hackman DA, Farah MJ, Meaney MJ (2010): Socioeconomic status
and the brain: Mechanistic insights from human and animal
research. Nat Rev Neurosci 11:651–659.
Hanson JL, Chandra A, Wolfe BL, Pollak SD (2011): Association
between income and the hippocampus. PloS One 6:e18712.
Hayasaka S, Phan KL, Liberzon I, Worsley KJ, Nichols TE (2004):
Nonstationary cluster-size inference with random field and
permutation methods. Neuroimage 22:676–687.
Hostinar CE, Sullivan RM, Gunnar MR (2014): Psychobiological
mechanisms underlying the social buffering of the
hypothalamic-pituitary-adrenocortical axis: A review of animal
models and human studies across development. Psychol Bull
140:256–282.
Javanbakht A, King AP, Evans GW, Swain JE, Angstadt M, Phan
KL, Liberzon I (2015): Childhood poverty predicts adult amyg-
dala and frontal activity and connectivity in response to emo-
tional faces. Front Behav Neurosci 9:
Jednorog K, Altarelli I, Monzalvo K, Fluss J, Dubois J, Billard C,
Dehaene-Lambertz G, Ramus F (2012): The influence of socioe-
conomic status on children’s brain structure. PloS One 7:
e42486.
K€uhn S, Gallinat J (2014): Segregating cognitive functions within hip-
pocampal formation: A quantitative meta-analysis on spatial navi-
gation and episodic memory. Human Brain Mapp 35:1129–1142.
Kim P, Evans GW, Angstadt M, Ho SS, Sripada CS, Swain JE,
Liberzon I, Phan KL (2013): Effects of childhood poverty and
chronic stress on emotion regulatory brain function in adult-
hood. Proc Natl Acad Sci 110:18442–18447.
Kong Xz, Song Y, Zhen Z, Liu J (2016): Genetic variation in S100B
modulates neural processing of visual scenes in Han Chinese.
Cerebral Cortex bhv322.
Kubarych TS, Prom-Wormley EC, Franz CE, Panizzon MS, Dale
AM, Fischl B, Eyler LT, Fennema-Notestine C, Grant MD,
Kremen WS (2012): A multivariate twin study of hippocampal
volume, self-esteem and well-being in middle-aged men.
Genes, Brain Behav 11:539–544.
Lawson GM, Duda JT, Avants BB, Wu J, Farah MJ (2013): Associa-
tions between children’s socioeconomic status and prefrontal
cortical thickness. Dev Sci 16:641–652.
Mantzicopoulos P (1990): Coping with school failure: Characteris-
tics of students employing successful and unsuccessful coping
strategies. Psychol Schools 27:138–143.
Marcoulides GA, Hershberger SL (1997). Multivariate Statistical
Methods: A First Course. Hillsdale, NJ: Erlbaum.
Martinelli P, Sperduti M, Piolino P (2013): Neural substrates of
the self-memory system: New insights from a meta-analysis.
Hum Brain Mapp 34:1515–1529.
McEwen BS (2012): Brain on stress: How the social environment
gets under the skin. Proc Natl Acad Sci 109:17180–17185.
McEwen BS, Gianaros PJ (2010): Central role of the brain in stress
and adaptation: Links to socioeconomic status, health, and dis-
ease. Ann N Y Acad Sci 1186:190–222.
r
 r Wang et al.
Mruk C (2006). Self-Esteem Research, Theory, and Practice:
Toward a Positive Psychology of Self-esteem, 3rd ed. New
York: Springer.
Nimon K (2010): Regression commonality analysis: Demonstration
of an SPSS solution. Multiple Linear Regression Viewpoints 36:
10–17.
Noble KG, Houston SM, Kan E, Sowell ER (2012): Neural corre-
lates of socioeconomic status in the developing human brain.
Dev Sci 15:516–527.
Noble KG, Houston SM, Brito NH, Bartsch H, Kan E, Kuperman
JM, Akshoomoff N, Amaral DG, Bloss CS, Libiger O (2015):
Family income, parental education and brain structure in chil-
dren and adolescents. Nat Neurosci 18:773–778.
Pan W, Liu C, Yang Q, Gu Y, Yin S, Chen A (2016): The neural
basis of trait self-esteem revealed by the amplitude of low-
frequency fluctuations and resting state functional connectiv-
ity. Soc Cognit Affect Neurosci 11:367–376.
Plotsky PM, Meaney MJ (1993): Early, postnatal experience alters
hypothalamic corticotropin-releasing factor (CRF) mRNA,
median eminence CRF content and stress-induced release in
adult rats. Mol Brain Res 18:195–200.
Pruessner JC, Baldwin MW, Dedovic K, Renwick R, Mahani NK,
Lord C, Lord C, Meaney M, Lupien S (2005): Self-esteem, locus
of control, hippocampal volume, and cortisol regulation in
young and old adulthood. Neuroimage 28:815–826.
Rosenberg, M. (1965). The measurement of self-esteem. In: Rosen-
berg M, editor. Society and the adolescent self image. New
York: Princeton University Press. pp. 297–307.
Sheridan MA, Sarsour K, Jutte D, D’Esposito M, Boyce WT (2012):
The impact of social disparity on prefrontal function in child-
hood. PloS One 7:e35744.
r 3766
r
Singh-Manoux A, Marmot MG, Adler NE (2005): Does subjective
social status predict health and change in health status better
than objective status? Psychosom Med 67:855–861.
Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens
TE, Johansen-Berg H, Bannister PR, De Luca M, Drobnjak I,
Flitney DE (2004): Advances in functional and structural MR
image analysis and implementation as FSL. Neuroimage 23:
S208–S219.
Somerville LH, Kelley WM, Heatherton TF (2010): Self-esteem
modulates medial prefrontal cortical responses to evaluative
social feedback. Cereb Cortex 20:3005–3013.
Song Y, Lu H, Hu S, Xu M, Li X, Liu J (2014). Regulating emotion
to improve physical health through the amygdala. Soc Cognit
Affect Neurosci 10: 523–530.
Staff RT, Murray AD, Ahearn TS, Mustafa N, Fox HC, Whalley LJ
(2012): Childhood socioeconomic status and adult brain size:
Childhood socioeconomic status influences adult hippocampal
size. Ann Neurol 71:653–660.
Taylor SE, Brown JD (1994): Positive illusions and well-being
revisited: Separating fact from fiction. Psychol Bull 116:21–28.
Wang M, Cai B, Wu Y, Dai X (2010): The factor structure of Chi-
nese Rosenberg’s Self-esteem Scale affected by item statement
method. Psychol Exploration 30:60–64.
Wang Y, Kong F, Huang L, Liu J: Neural correlates of biased
responses: The negative method effect in the Rosenberg Self-
Esteem Scale is associated with right amygdala volume. J Pers
(in press).
Werner E (1995): Resilience in development. Curr Direct Psychol
Sci 4:81–85.
Werner E (1997): Vulnerable but invincible: High-risk children
from birth to adulthood. Acta Paediatr 422:103–105.
Yang Y, Wang D (2007): Retest the bidimensional model of Rosen-
berg Self-esteem Scale. Chinese Mental Health J 21:603–609.
r