44 doi: 10.1111/j.1365-3016.2010.01161.

Duration of breast feeding and language ability in

middle childhood ppe_1161 44..52

Andrew J. O. Whitehousea,b, Monique Robinsona, Jianghong Lia,c and Wendy H. Oddya

a
Telethon Institute for Child Health Research, Centre for Child Health Research, and bNeurocognitive Development Unit, School of Psychology,
The University of Western Australia, and cCentre for Developmental Health, Curtin Health Innovation Research Institute, Curtin University of
Technology, Perth, Western Australia, Australia

Summary
Correspondence: Whitehouse AJO, Robinson M, Li J, Oddy WH. Duration of Breast feeding and lan-
Dr Andrew Whitehouse, guage ability in middle childhood. Paediatric and Perinatal Epidemiology 2011; 25: 44–52.
Telethon Institute for Child
Health Research, Centre for There is controversy over whether increased breast-feeding duration has long-term
Child Health Research, The benefits for language development. The current study examined whether the positive
University of Western associations of breast feeding on language ability at age 5 years in the Western Australian
Australia, West Perth,
Pregnancy (Raine) Cohort, were still present at age 10 years. The Raine Study is a
Western Australia 6872,
Australia. longitudinal study of 2868 liveborn children recruited at approximately 18 weeks
E-mail: gestation. Breast-feeding data were based upon information prospectively collected
awhitehouse@ichr.uwa.edu.au during infancy, and were summarised according to four categories of breast-feeding
duration: (1) never breast-fed, (2) breast-fed predominantly for <4 months, (3) breast-fed
predominantly for 4–6 months, and (4) breast-fed predominantly for >6 months. Lan-
guage ability was assessed in 1195 children at the 10 year follow-up (mean age = 10.58
years; standard deviation = 0.19) using the Peabody Picture Vocabulary Test – Revised
(PPVT-R), which is based around a mean of 100 and a standard deviation of 15.
Associations between breast-feeding duration and PPVT-R scores were assessed before
and after adjustment for a range of sociodemographic, obstetric and psychosocial
covariates. Analysis of variance revealed a strong positive association between the
duration of predominant breast feeding and PPVT-R at age 10 years. A multivariable
linear regression analysis adjusted for covariates and found that children who were
predominantly breast-fed for >6 months had a mean PPVT-R score that was 4.04 points
higher than children who were never breast-fed. This compared with an increase of 3.56
points at age 5 years. Breast feeding for longer periods in early life has a positive and
statistically-independent effect on language development in middle childhood.

Keywords: breast feeding, language development, Raine Study.

Introduction
studies, with several investigations highlighting a
While language development in humans is under clear dose–response relationship, with greater effects
substantial genetic control,1,2 there is accumulating on language associated with longer breast-feeding
evidence that environmental factors can alter the duration,4–7 and others reporting a negligible effect.8–11
expression of the genome and lead to enduring phe- One possible explanation for the inconsistent
notypic changes. Breast feeding is one environmental findings relates to the methodological limitations of
factor known to confer a range of nutritional and observational studies. Mothers who breast feed tend
immunological advantages upon infants,3 and there to have higher levels of education and a more stable
has been considerable interest in whether breast milk home environment.12 A failure to adjust for these
may also benefit language development in children. variables may overestimate the effects of breast
However, findings have been inconsistent across feeding.

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
 Breast feeding and language ability 45

Our study design, with prospectively collected data KEMH and/or Princess Margaret Hospital for Chil-
on breast feeding, longitudinal follow-up of language dren in Perth.
development and a wide range of data on potential From the 2900 pregnancies recruited into the Raine
confounders, has unique potential for determining study, 2868 liveborn children have been comprehen-
the relationship between breast feeding and language sively phenotyped since the commencement of the
development with comprehensive adjustment for study. The current investigation is based on data that
environmental confounders. We have previously were collected in the follow-up assessments at ages 1, 2,
investigated the effects of a longer breast-feeding 3, 5 and 10 years. Because the current study had an
duration and verbal IQ scores at age 5 years.13 That interest in the language outcomes of the birth cohort,
study found that in early childhood (mean age = 5.9 children whose ability may be compromised because of
years, SD = 0.2, range = 5.2–6.8), there was a positive speaking a language other than English at home were
association between the duration of predominant excluded from analyses (n = 163), as were children who
breast feeding in infancy and verbal IQ. Children had Down’s syndrome, autism or intellectual disability
who were predominantly breast-fed for >6 months with a known cause (n = 37), and children of Aboriginal
had mean verbal IQ scores [(indexed by the Peabody mothers because of missing language data at the 10-year
Picture Vocabulary Test – Revised (PPVT-R)] 6.44 follow-up (n = 67). The small number of preterm infants
points higher than children who were never breast- in the Raine cohort (n = 179) prevented any meaningful
fed. Following adjustment for perinatal, social and analysis of these individuals in our previous investiga-
family factors, breast feeding for >6 months tion,13 and therefore these individuals were also
was associated with an increase in verbal IQ of 3.56 excluded from the current analysis.
points. There were 1976 participants followed up at age 10
The current study provides a longitudinal follow-up years for whom complete data on breast feeding were
investigation of these previous findings on breast available. The longitudinal attrition rate was in keeping
feeding and language ability among children in the with other cohort studies,18 with around 60% of these
Western Australian Pregnancy Cohort (Raine) Study. participants (n = 1195) providing language data at the
The current study seeks to determine whether the asso- 10 year follow-up. The mean age was 10.58 years
ciation between breast feeding and language ability (SD = 0.19; range = 9.39–12.37), with 1155 out of 1195
persists until middle childhood, and the magnitude of children (97%) assessed between their tenth and elev-
any effect. enth birthdays.

Method Predictor variable

Data on the duration of breast feeding (in months) was
Participants
collated from feeding questionnaires completed by
The Raine Study is a longitudinal study of women and mothers at the 1-, 2- and 3-year follow-ups. Predomi-
their children recruited at approximately 18 weeks of nant breast feeding was presumed to occur up to the
pregnancy from the public antenatal clinic at King introduction of milk other than breast milk. This defi-
Edward Memorial Hospital (KEMH) or surrounding nition did not preclude the intake of solid foods. In
private clinics in Perth (Australia). Approximately 100 accordance with our previous investigation, a 4-tier
women per month were enrolled from August 1989 to categorical variable was created: (1) never breast-fed,
April 1992, with a final sample of 2804 pregnancies. (2) breast-fed predominantly for <4 months, (3) breast-
The inclusion criteria were a gestational age between fed predominantly for 4–6 months, and (4) breast-fed
16 and 20 weeks (average 18 weeks), English language predominantly for >6 months. A similar 4-tiered vari-
skills sufficient to understand the study demands, an able representing the duration of any breast feeding
expectation to deliver at KEMH, and an intention to was also generated.
remain in Western Australia to enable future follow-up
of their child. Full details of enrolment methods are
Outcome variable
included in Newnham and colleagues.14 Participant
recruitment and all follow-ups of the study families At both the 5- and 10-year follow-ups, children com-
were approved by the Human Ethics Committee at pleted the PPVT-R,15 which is a test of receptive

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
46 A. J. O. Whitehouse et al.

vocabulary requiring participants to choose which of Statistical analyses

four pictures corresponds to a word presented orally
Statistical analyses followed four stages. First, prelimi-
by the experimenter. Raw scores are transformed to
nary analyses were undertaken to investigate the
standard scores, based around a mean of 100 and a SD
pattern of sample attrition. Second, frequency distri-
of 15.
butions for all predictor and control variables were
examined. Third, PPVT-R scores across the four
breast-feeding groups were compared. Finally, a mul-
Covariates tivariable regression analysis was conducted using a
Covariates previously found to influence child devel- general linear model to examine the effect of the cat-
opment in the Raine cohort16 were also considered. egorical breast-feeding variable and covariates on con-
These variables included: (1) sociodemographic factors tinuous PPVT-R scores. Missing data were deleted
recorded at 18 weeks pregnancy (maternal age at con- casewise. All statistical analyses were undertaken
ception, maternal education, family income, and the using SPSS software version 15.0 (SPSS Inc., Chicago,
presence of the biological father in the family home), IL, USA). The analyses reported in the current study
(2) antenatal variables recorded at 34 weeks pregnancy were repeated exchanging the ‘any breast-feeding’
(maternal smoking and alcohol consumption during variable for the ‘predominant breast-feeding’ variable,
pregnancy, maternal experience of stressful events and an identical pattern and magnitude of findings
during pregnancy), and (3) obstetric variables recorded was observed. Therefore, we focused on the duration
at birth (parity, gestational age, child’s sex, and propor- of predominant breast feeding for this study.
tion of optimal birthweight, a measure of the appro-
priateness of fetal growth17).
A range of postnatal variables were also considered.
Results
Maternal experience of stressful events during the first
Comparability of samples at age 5 and age
year of the infant’s life was recorded at the 1-year
10 years
follow-up. Family functioning was measured using the
General Functioning Scale from the McMaster Family There was a strong positive association between
Assessment Device (MMFAD)18 at the 3-year follow- PPVT-R scores at age 5 and 10 for the 1067 children
up. This short-form scale consists of 13 statements who provided data at both time points, r = 0.60,
derived from an item-analysis of the complete 60-item P < 0.001. Differences in the sociodemographic, obstet-
scale, and includes questions on family communica- ric, antenatal and postnatal characteristics between
tion, problem solving, affective responsiveness and those participants who contributed language data at
behaviour control. Scores range from 0–39, with higher the 5-year but not the 10-year follow-up, and those
scores indicating more cohesive family functioning. who contributed language data at both time points
Five-year MMFAD data were used if 3-year data were (and for whom breast-feeding data were available)
missing. The quality of the language-learning environ- were investigated with independent samples t-tests
ment was measured at the 3-year follow-up by asking (continuous variables) or chi-squared analyses (cat-
parents how often they read to their child. Responses egorical variables). Children who provided PPVT-R
were grouped into four categories: ⱕonce a week, 2–3 data at both time points had mothers who were: older
times a week, once a day, or >once a day. Finally, at the (language data at age 5 and 10 years: mean = 28.70,
10-year follow-up, parents completed the Parenting SD = 5.73; language data at age 5 years only:
Scale,19 a 30-item questionnaire that measures dysfunc- mean = 27.21, SD = 6.14), more likely to live with the
tional discipline styles. Three subscales are derived child’s biological father during pregnancy (89.5%;
from this questionnaire: laxness (permissive disci- 85.1%), less likely to smoke 11+ cigarettes per day
pline), over-reactivity (authoritarian discipline, dis- during pregnancy (8.9%; 13.6%), and experienced
plays of anger, meanness and irritability) and fewer stressful events during the first year of the
verbosity (overly long reprimands or reliance on child’s life (mean = 1.44, SD = 1.36; mean = 1.62,
talking) measured on a 7-point Likert scale. Data were SD = 1.51). Participants who provided data at both the
coded so that higher scores indicated a more dysfunc- 5- and 10-year follow-up had significantly lower
tional disciplinary style. PPVT-R scores at 5 years of age (mean = 106.4,

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
 Breast feeding and language ability 47

SD = 12.6) than children who provided data at the Effect of breast feeding on language ability at age
5-year follow-up only (mean = 108.7, SD = 15.2). All of 10 years
these differences were significant at P < 0.05. However,
The frequency characteristics of the cohort who had
chi-squared analyses found no difference between the
breast feeding and language data at age 10 years
two groups in the distribution of the 4-tiered breast-
(n = 1195) are presented in Table 1. Approximately 50%
feeding variable, P = 0.923.
Table 1. Frequency distribution of the predictor variable (breast-feeding duration) and covariates (other factors) for children whose
language and breast feeding data were available at age 10 years

Continuous variables n Mean (SD)

Maternal age at conception (years) 1195 28.43 (5.72)

Stressful events in pregnancy (number of events) 1153 1.79 (1.78)
Gestational age at delivery (weeks) 1195 39.61 (1.31)
Proportion of expected birthweight (%) 1193 98.04 (11.95)
5 min Apgar scores 1192 9.04 (0.65)
Year one stressful events (number of events) 1179 1.46 (1.35)
McMaster Family Assessment Device 1127 20.58 (2.49)
Parenting Scale
Laxness 1151 2.83 (0.68)
Over-reactivity 1153 2.79 (0.84)
Verbosity 1149 4.26 (0.82)

Categorical variables N n (%)

Maternal education at pregnancy 1195

Completed secondary school 507 (42.4)
Did not complete secondary school 688 (57.6)
Smoking in pregnancy 1153
None 912 (79.1)
1–10 cigarettes daily 137 (11.9)
11+ cigarettes daily 104 (9.0)
Alcohol consumption during pregnancy 1153
None 678 (58.8)
Once a week or less 404 (35.0)
Several times a week or more 71 (6.2)
Family income during pregnancy 1151
ⱖ$24,000 731 (63.5)
<$24,000 420 (36.5)
Biological father living with family during pregnancy 1190
Yes 1068 (89.8)
No 122 (10.3)
Sex of child 1195
Male 627 (52.5)
Female 568 (47.5)
Parity 1195
0 529 (44.3)
ⱖ1 666 (55.7)
Reading to child at age 3 years 1104
ⱕOnce a week 116 (10.5)
2 or 3 times per week 321 (29.1)
Once a day 388 (35.1)
>Once a day 279 (25.3)
Predominant breast-feeding duration 1195
Never breast-fed 105 (8.8)
<4 months 443 (37.1)
4–6 months 309 (25.9)
>6 months 338 (28.3)

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
48 A. J. O. Whitehouse et al.

Table 2. Peabody Picture Vocabulary Test – Revised15 scores at age 10 years for the four groups representing different durations of
predominant breast feeding

Predominant breast feeding

Never breast-fed <4 months 4–6 months >6 months

n 105 443 308 338

Mean (SD) 100.9 (11.63) 103.46 (11.69) 106.44 (11.22) 107.13 (11.77)

of the sample was predominantly breast-fed for ⱖ4 smoking of 11+ cigarettes per day during pregnancy
months, with just over half of these children (28.3% of and having at least one older sibling were significantly
the total sample) predominantly breast-fed for >6 associated with poorer language scores at age 10 years.
months. Less than 10% of the sample was never pre- A progressively stronger positive association was
dominantly breast-fed.
A one-way anova revealed a significant difference Table 3. Adjusted multivariable linear regression models
between the PPVT-R scores of the groups representing showing the effect of the duration of predominant breast feeding
different predominant breast-feeding durations, on receptive vocabulary at age 10 years
F(3,1193) = 12.32, P < 0.001 (Table 2). Bonferroni post-
b [95% confidence
hoc tests found that the children who had been pre-
interval] P
dominantly breast-fed for between 4 and 6 months or
for >6 months, had greater language scores than chil- Covariatesa
dren who had been predominantly breast-fed for 0–4 Maternal age at conception 0.28 [0.14, 0.42] <0.001
months or not at all (for all comparisons, P < 0.05). Smoking during pregnancy
None 0.00 Reference 0.002
Univariable linear regression analysis found no rela-
1–10 cigarettes daily -1.62 [-3.74, 0.5]
tion between language and gestational age at delivery,
11+ cigarettes daily -3.16 [-5.56, -0.77]
proportion of optimal birthweight, stressful events Alcohol during pregnancy
during pregnancy or the first year of life, Apgar scores Not at all 0.00 Reference 0.001
at 5 min, low family income, cohabitation with the bio- ⱕOnce a week 1.8 [0.36, 3.24]
logical father, the verbosity subscale of the parenting >Once a week 3.53 [0.56, 6.51]
Low maternal education -2.27 [-3.72, -0.83] 0.002
questionnaire, and PPVT-R scores at age 10 years (for
Male gender 1.49 [0.17, 2.82] 0.027
all analyses, P > 0.3). These variables were therefore Parity of ⱖ1 -2.4 [-3.87, -0.92] 0.001
excluded from further analysis. Table 3 presents the McMaster Family Assessment 0.32 [0.04, 0.59] 0.024
results of the multivariable regression analyses exam- Device at age 3 years
ining the effect of the predictor variables and covariates Reading to child
ⱕOnce a week 0.00 Reference 0.043
on language outcome. There were significant associa-
2 or 3 times per week 2.83 [0.42, 5.24]
tions between the duration of predominant breast
Once a day 2.9 [0.5, 5.3]
feeding and language ability at age 10 years. Children >Once a day 6.41 [3.83, 9]
who had been predominantly breast-fed for <4 months Parenting style at age 10 years
had higher language scores than children who had Lax -0.46 [-1.53, 0.62] 0.407
never been breast-fed [(regression coefficient Over-reactive -0.27 [-0.61, 1.14] 0.550
Predictor variable of interest
(b) = 2.71)], while the effect was stronger for children
Duration of predominant
predominantly breast-fed for between 4 and 6 months breast feeding
(b = 3.83), and stronger still for children who were pre- Not at all 0.00 Reference 0.001
dominantly breast-fed for >6 months (b = 4.04). <4 months 2.71 [0.23, 5.18]
Of the potential confounders included in the model 4 to 6 months 3.83 [1.24, 6.42]
>6 months 4.04 [1.48, 6.6]
(Table 3), a higher maternal age at conception, male
gender and low levels of alcohol during pregnancy a
Significant covariates only are presented.
were associated with small increases in language b, regression coefficient; MMFAD, McMaster Family Assessment
scores. Lower levels of maternal education, maternal Device.

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
 Breast feeding and language ability 49

observed between the frequency of reading to the child least equivalent magnitude to the association in early
at age 3 years and year 10 PPVT-R scores. There was no childhood.
association between a lax or overreactive parenting Previous investigations of the association between
style and language ability in middle childhood. breast-feeding duration and language development
have had mixed findings. One possible explanation for
the inconsistency is the lack of control for potentially
confounding variables. Meta-analyses suggest that
Discussion
<25% of studies in this area have adjusted for sociode-
The current study demonstrates a positive effect of mographic confounders.23 Mothers who breast feed
increased breast-feeding duration on language ability tend to have higher levels of education and more posi-
in children at 10 years of age. Simple comparisons tive home environments,12,21 and a recent multisite
demonstrated progressively higher language scores for study found that a positive effect of breast feeding on
children who were predominantly breast-fed for language ability at 3 years of age (assessed using the
longer during infancy. This effect remained after PPVT – 3rd edition) was observed only among children
adjustment for a range of potential covariates, includ- whose mothers also engaged in positive parenting
ing maternal education and the quality of the family behaviours.24 While the current study did not include
and language-learning environment, suggesting that direct measurement of positive parenting practices, the
the association was not merely reflective of sociodemo- regression model was adjusted for indices of family
graphic advantage. functioning (MMFAD), parent discipline (Parenting
The current findings extend those of a previous Scale) and child engagement (time spent reading to the
investigation of the Raine cohort,13 and have implica- child), all of which are likely to bear considerable influ-
tions for our understanding of the longitudinal effects ence on the overall parent–child relationship. A second
of breast feeding on language development. At age 5 possibility is that the mixed findings reflect a variable
years, children who were predominantly breast-fed for effect of breast feeding on language ability over time.
>6 months had an adjusted increase in PPVT-R scores For example, it is possible that breast feeding advan-
of 3.56 points compared with children who were never tages early language development, but that this effect is
breast-fed. The comparable increase in PPVT-R score at reduced or extinguished as children commence school
age 10 years was 4.04 points. and receive more formal language instruction. Existing
An important issue in the interpretation of the studies provide some evidence for this proposal, with
current findings is sample attrition. A non-random null effects commonly reported among children over
pattern of attrition is commonly observed in longitudi- the age of 7 years.8–10,25 With our current investigation,
nal cohort studies,20 and the current study found that we were in a relatively unique position to test both of
socially disadvantaged women and their children were these possibilities, adjusting for a wide range of envi-
less likely to participate at the 10-year follow-up. While ronmental confounders and examining the same
the loss of these participants may have affected the cohort at different time points using the same language
results of this study, the original cohort over- measure. The finding that breast-feeding duration
represented socially disadvantaged women,21 and improves language ability in middle childhood sug-
therefore this pattern of attrition may have increased gests that this association is independent of numerous
the extent to which the remaining cohort is represen- covariates and is evident well beyond the commence-
tative of the general population and hence the findings ment of formal education.
can be generalised. Furthermore, computer simula- The mechanisms underpinning the association
tions using data from the Avon Longitudinal Study of between breast-feeding duration and language devel-
Parents and Children (UK), have found that selective opment are likely to be multifactorial. One proposal is
dropout in cohort studies only marginally affects that breast milk contains several nutrients not found in
regression coefficients if participant selection occurs formula milk that are essential for optimum brain
according to the predictor variables (as was the case in growth, such as long-chain polyunsaturated fatty acids
the current study).22 Based on these data, we suggest (LCPUFA).26 LCPUFAs are necessary structural ele-
that the current findings provide preliminary evidence ments of cell membranes and are therefore essential in
that the beneficial effect of breast feeding on the formation of new tissue, including neurons.24
language development in middle childhood is of at Animal studies have found that LCPUFA may also

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
50 A. J. O. Whitehouse et al.

play a neuroprotective role in early development, important to note that the PPVT-R is highly correlated
eliminating the build up of certain molecules to neuro- with other widely used indices of language ability,38
toxic levels (e.g. calcium ions) and inhibiting such as verbal IQ measured by the Wechsler Intelli-
glutamatergic synaptic transmission.27 Further stud- gence Scale for Children. Second, while we controlled
ies28,29 have identified that some LCPUFAs, including for a range of factors that may generate individual
eicosapentaenoic acid, docosahexaenoic and arachi- variation in language ability, our study did not provide
donic acid, act as secondary messengers, inducing the a direct measure of parental language skills. Language
release of acetylcholine and noradrenaline, which are ability is known to be moderately heritable,2 and one
neurotransmitters known to be involved in cognitive study has found that once maternal IQ has been taken
development.30,31 A second proposal is that breast into account, there is no significant association
feeding has a positive effect on the mother–child rela- between breast feeding and early cognitive and lan-
tionship, thereby facilitating bonding, interaction and, guage development.39 However, this finding has not
indirectly, cognitive growth.6 There is also evidence been replicated in studies of other cohorts, which have
that positive maternal contact may induce a biological identified an increase in child cognitive and language
response in the offspring. In rats, variations in maternal ability with breast feeding that is independent of
licking and grooming have been shown to promote the parental IQ.40,41 Furthermore, a large randomised con-
development of neural systems that mediate trolled trial of a breast-feeding promotion intervention,
hypothalamic–pituitary–adrenal axis and behavioural in which the level of parental IQ should be distributed
responses to stress in the offspring as well as certain randomly between treatment and control groups, iden-
forms of learning and memory.32,33 However, no tified a 7.5 point advantage on the WISC verbal IQ for
research has examined these associations in humans. children who were breast-fed for a longer duration.42
In accordance with previous research,34 maternal In summary, our study identified a dose–response
cigarette smoking during pregnancy was found to have relationship between the duration of predominant
a deleterious effect on the offspring’s language ability breast feeding and higher language scores at 10 years
at age 10 years. However, an unexpected finding was of age. The magnitude of this association was compa-
that maternal alcohol consumption during pregnancy rable to the effect observed in the same cohort in early
was associated with increased PPVT-R scores in chil- childhood and indicates long-term advantages of
dren. This finding is consistent with recent data breast feeding that may have biological or psychologi-
showing that low to moderate levels of alcohol intake cal origins or both.
during pregnancy may not be associated with poor
developmental outcome in offspring.35,36 However, we
would like to emphasise that in the current study we
Acknowledgements
were unable to identify those mothers who drank
heavily during pregnancy (11+ drinks per week), and The authors would like to acknowledge the National
that there are consistent findings across multiple Health and Medical Research Council (NH&MRC) for
studies that high levels of alcohol exposure during their long-term contribution to funding the study over
pregnancy are associated with an increase in adverse the last 20 years. Core Management of the Raine study
outcomes for the offspring.37 has been funded by the University of Western Austra-
The main strength of the current study design was lia (UWA), the UWA Faculty of Medicine, Dentistry
the prospectively assessed community-based sample, and Health Sciences, the Raine Medical Research
which enabled a statistically powerful examination of Foundation, the Telethon Institute for Child Health
the longitudinal association between breast feeding Research, and the Women’s and Infants Research
and language development in middle childhood. Foundation. The 1-, 2-, 3- and 10-year follow-ups were
However, several limitations of the current design also funded by the NH&MRC and the Raine Medical
deserve noting. First, the assessment of language Research Foundation. The authors are extremely grate-
ability was limited to one psychometric test. While this ful to the study participants and their families as well
enabled us to reduce (but not eliminate) measurement as the Raine Study team for cohort co-ordination and
error between time points, the PPVT-R assesses recep- data collection. The authors also acknowledge Emma
tive vocabulary only, and does not provide any Jaquet, who made helpful comments on a previous
measure of expressive language skills. However, it is version of this manuscript.

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
 Breast feeding and language ability 51

References 18 Epstein N, Baldwin L, Bishop D. The McMaster Family

Assessment Device. Journal of Marital and Family Therapy
1 Spinath FM, Price TS, Dale PS, Plomin R. The genetic and 1983; 9:171–180.
environmental origins of language disability and ability. 19 Arnold DS, O’Leary SG, Wolff LS, Acker MM. The Parenting
Child Development 2004; 75:445–454. Scale: a measure of dysfunctional parenting in discipline
2 Stromswold K. The heritability of language: a review and situations. Psychological Assessment 1993; 5:137–144.
metaanalysis of twin, adoption, and linkage studies. 20 O’Connor TG, Heron J, Golding J, Glover V, ALSPAC Study
Language 2001; 77:647–723. Team. Maternal antenatal anxiety and behavioural/
3 Stanner S, Smith E. Breastfeeding: early influences on later emotional problems in children: a test of a programming
health. Nutrition Bulletin 2005; 30:94–102. hypothesis. Journal of Child Psychology and Psychiatry, and
4 Jacobson SW, Chiodo LM, Jacobson JL. Breastfeeding effects Allied Disciplines 2003; 44:1025–1036.
on intelligence quotient in 4- and 11-year-old children. 21 Li J, Kendall GE, Henderson S, Downie J, Landsborough L,
Pediatrics 1999; 103:e711–e716. Oddy WH. Maternal psychosocial well-being in pregnancy
5 Niemala A, Jarvenpaa A-L. Is breastfeeding beneficial and and breastfeeding duration. Acta Paediatrica 2008;
maternal smoking harmful to the cognitive development of 97:221–225.
children? Acta Paediatrica 1996; 85:1202–1206. 22 Wolke D, Waylen A, Samara M, Steer C, Goodman R, Ford
6 Quinn PJ, O’Callaghan M, Williams GM, Najman JM, T, et al. Selective drop-out in longitudinal studies and
Andersen MJ, Bor W. The effect of breastfeeding on child non-biased prediction of behaviour disorders. The British
development at 5 years: a cohort study. Journal of Paediatrics Journal of Psychiatry 2009; 195:249–256.
and Child Health 2001; 37:465–469. 23 Jain A, Concato J, Leventhal JM. How good is the evidence
7 Taylor B, Wadsworth J. Breast-feeding and linking breastfeeding and intelligence? Pediatrics 2002;
child-development at 5 years. Developmental Medicine and 109:1044–1053.
Child Neurology 1984; 26:73–80. 24 Gibson-Davis CM, Brooks-Gunn J. Breastfeeding and verbal
8 Doyle LW, Rickards AL, Kelly EA, Ford GW, Callanan C. ability of 3-year-olds in a multicity sample. Pediatrics 2006;
Breast-feeding and intelligence. Lancet 1992; 339:744–745. 118:e1444–e1451.
9 Fergusson DM, Beautrais AL, Silva PA. Breast-feeding and 25 Wigg NR, Tong SL, McMichael AJ, Baghurst PA, Vimpani G,
cognitive-development in the first 7 years of life. Social Roberts R. Does breastfeeding at six months predict
Science & Medicine 1982; 16:1705–1708. cognitive development? Australian and New Zealand Journal
10 Malloy MH, Berendes H. Does breast-feeding influence of Public Health 1998; 22:232–236.
intelligence quotients at 9 and 10 years of age? Early Human 26 Larque E, Demmelmair H, Koletzko B. Perinatal supply and
Development 1998; 50:209–217. metabolism of long-chain polyunsaturated fatty acids.
11 Silva PA, Buckfield P, Spears GF. Some maternal and child Annals of the New York Academy of Sciences 2002; 967:299–310.
developmental characteristics associated with breast 27 Lauritzen I. Polyunsaturated fatty acids are potent
feeding: report from the Dunedin Multidisciplinary neuroprotectors. EMBO Journal 2000; 19:1784.
Child-Development Study. Australian Paediatric Journal 1978; 28 Das UN. Long-chain polyunsaturated fatty acids in the
14:265–268. growth and development of the brain and memory.
12 Michaelsen KF, Larsen PS, Thomsen BL, Samuelson G. The Nutrition 2003; 19:62–65.
Copenhagen Cohort Study on infant nutrition and growth – 29 Farkas E, de Wilde MC, Kiliaan AJ, Meijer J, Keijser JN,
Duration of breast-feeding and influencing factors. Acta Luiten PGM. Dietary long chain PUFAs differentially affect
Paediatrica 1994; 83:565–571. hippocampal muscarinic 1 and serotonergic 1A receptors in
13 Oddy WH, Kendall GE, Blair E, de Klerk NH, Stanley FJ, experimental cerebral hypoperfusion. Brain Research 2002;
Landau LI, et al. Breast feeding and cognitive development 954:32–41.
in childhood: a prospective birth cohort study. Paediatric and 30 Dalley JW, McGaughy J, O’Connell MT, Cardinal RN, Levita
Perinatal Epidemiology 2003; 17:81–90. L, Robbins TW. Distinct changes in cortical acetylcholine
14 Newnham JP, Evans SF, Michael CA, Stanley FJ, Landau LI. and noradrenaline efflux during contingent and
Effects of frequent ultrasound during pregnancy: a noncontingent performance of a visual attentional task.
randomized controlled trial. Lancet 1993; 342:887–891. Journal of Neuroscience 2001; 21:4908–4914.
15 Dunn L, Dunn L. Peabody Picture Vocabulary Test-Revised: 31 Sarter M, Bruno JP. Cognitive functions of cortical
Manual. Circle Pines, MN: American Guidance Services, acetylcholine: toward a unifying hypothesis. Brain Research
1981. Reviews 1997; 23:28–46.
16 Robinson M, Oddy WH, Li JH, Kendall GE, de Klerk NH, 32 Caldji C, Tannenbaum B, Sharma S, Francis D, Plotsky PM,
Silburn SR, et al. Pre- and postnatal influences on preschool Meaney MJ. Maternal care during infancy regulates the
mental health: a large-scale cohort study. Journal of Child development of neural systems mediating the expression of
Psychology and Psychiatry 2008; 49:1118–1128. fearfulness in the rat. Proceedings of the National Academy of
17 Blair E, Liu Y, de Klerk N, Lawrence D. Optimal fetal Sciences of the United States of America 1998; 95:5335–5340.
growth for the Caucasian singleton and assessment of 33 Liu D, Diorio J, Day JC, Francis DD, Meaney MJ. Maternal
appropriateness of fetal growth: an analysis of a total care, hippocampal synaptogenesis and cognitive
population perinatal database. BMC Pediatrics 2005; development in rats. Nature Neuroscience 2000;
5:13. 3:799.

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.
52 A. J. O. Whitehouse et al.

34 Lassen K, Oei TPS. Effects of maternal cigarette smoking analysis, and meta-analysis. British Medical Journal 2006;
during pregnancy on long-term physical and cognitive 333:945–948.
parameters of child development. Addictive Behaviors 1998; 39 Breen MJ. Comparison of the Wechsler Intelligence Scale for
23:635–653. Children Revised and the Peabody Picture-Vocabulary
35 Kelly Y, Sacker A, Gray R, Kelly J, Wolke D, Quigley MA. Test-Revised for a referred population. Psychological Reports
Light drinking in pregnancy, a risk for behavioural 1981; 49:717–718.
problems and cognitive deficits at 3 years of age? 40 Gómez-Sanchiz M, Cañete R, Rodero I, Baeza JE, González
International Journal of Epidemiology 2009; 38:129–140. JA. Influence of breast-feeding and parental intelligence on
36 Robinson M, Oddy WH, McLean NJ, Jacoby P, Pennell CE, cognitive development in the 24-month-old child. Clinical
de Klerk NH et al. Low–moderate prenatal alcohol exposure Pediatrics 2004; 43:753–761.
and risk to child behavioural development: a prospective 41 Angelsen NK, Vik T, Jacobsen G, Bakketeig LS. Breast
cohort study. British Journal of Obstetrics and Gynaecology feeding and cognitive development at age 1 and
2010; 117:1139–1150. 5 years. Archives of Disease in Childhood 2001; 85:183–
37 Leary CMO. Fetal alcohol syndrome: diagnosis, 188.
epidemiology, and developmental outcomes. Journal of 42 Kramer MS, Aboud F, Mironova E, Vanilovich I, Platt RW,
Paediatrics and Child Health 2004; 40:2–7. Matush L, et al. Breastfeeding and child cognitive
38 Der G, Batty GD, Deary IJ. Effect of breast feeding on development: new evidence from a large randomized trial.
intelligence in children: prospective study, sibling pairs Archives of General Psychiatry 2008; 65:578–584.

Book review

Ethics and Epidemiology, Second Edition. Edited by articulated, refreshingly free from jargon and fre-
S. S. Coughlin, T. L. Beauchamp and D. L. Weed. Oxford: quently put into significant historical context. Contro-
Oxford University Press, 2009, pp. 315, £40. ISBN 978- versial positions are not evaded and in one chapter, the
0-19-532293-4. reader is asked to consider ‘the universal adoption of
the practice of full disclosure’ in regard to the disclo-
As the editors acknowledge, although this second sure of individual test results. Even the section one
edition of Ethics and Epidemiology is ‘most suited to the might assume to be most culturally specific, ‘Regula-
situation in North America . . . much of the book’s tory Context and Professional Education’, provides
content can easily be construed beyond the North interest and stimulation for researchers working in any
American context’ and therefore it will be of great international location. The one chapter not written by
value to epidemiological researchers anywhere in the academics based in North America but by UK academ-
world requiring easily accessible analyses of generic ics provides an excellent discussion of the sometimes
ethical issues. Not only are familiar topics covered complex ethical issues in international health research.
such as consent, confidentiality, risks and benefits but For all those struggling to keep abreast of the
also more recent dilemmas from this fast moving dis- ethical standards and dilemmas in current epidemio-
cipline, for example, Genome Wide Association logical practice, this comprehensive volume is highly
Studies and subsequent data sharing, globalisation recommended.
and internet research. The discussions are clearly Karen Birmingham

Paediatric and Perinatal Epidemiology, 25, 44–52. © 2010 Blackwell Publishing Ltd.