861436

research-article2019
PSSXXX10.1177/0956797619861436Havron et al.The Effect of Older Siblings on Language Development

ASSOCIATION FOR
Research Article PSYCHOLOGICAL SCIENCE
Psychological Science

The Effect of Older Siblings on Language 1­–11

© The Author(s) 2019
Article reuse guidelines:
Development as a Function of Age sagepub.com/journals-permissions
DOI: 10.1177/0956797619861436
https://doi.org/10.1177/0956797619861436

Difference and Sex www.psychologicalscience.org/PS

TC

Naomi Havron1 , Franck Ramus1, Barbara Heude2,

Anne Forhan2, Alejandrina Cristia1, Hugo Peyre1,2,3, and
the EDEN Mother-Child Cohort Study Group
1
Laboratoire de Sciences Cognitives et Psycholinguistique, Département d’Etudes Cognitives, École
Normale Supérieure, École des Hautes Etudes en Sciences Sociales, Centre Nationale de la Recherche
Scientifique, PSL Université; 2Université de Paris, CRESS Centre of Research in Epidemiology and Statistics,
Institut National de la Santé et de la Recherche Médicale, Institut National de la Recherche Agronomique,
France; and 3Department of Child and Adolescent Psychiatry, Robert Debré Hospital, Paris, France

Abstract
The number of older siblings a child has is negatively correlated with the child’s verbal skills, perhaps because
of competition for parents’ attention. In the current study, we examined the role of siblings’ sex and age gap as
moderating factors, reasoning that they affect older siblings’ tendency to compensate for reduced parental attention.
We hypothesized that children with an older sister have better language abilities than children with an older brother,
especially when there is a large age gap between the two siblings. We reanalyzed data from the EDEN cohort (N =
1,154) and found that children with an older sister had better language skills than those with an older brother. Contrary
to predictions, results showed that the age gap between siblings was not associated with language skills and did not
interact with sex. Results suggest that the negative effect of older siblings on language development may be entirely
due to the role of older brothers. Our findings invite further research on the mechanisms involved in this effect.

Keywords
language, language development, siblings, sex differences, preregistered

Received 1/23/19; Revision accepted 5/24/19

The number of older siblings a child has is negatively down. Thus, having more siblings is almost always det-
correlated with that child’s verbal skills, including ver- rimental, but the older the siblings, the less detrimental
bal intelligence, language-development measures, and their effect.
educational attainment (e.g., Black, Devereux, & Verbal intelligence is more affected by birth order
Salvanes, 2005; Peyre, Bernard, et al., 2016). According than is nonverbal intelligence (Peyre, Bernard, et al.,
to the resource-dilution model (Blake, 1981), a family 2016), suggesting that the effect of older siblings is
has limited resources to distribute among siblings, and greater for language development. One possible reason
the more children the family has, the less resources are for this may be that language development is more
allocated to each child. These resources can be material affected by the quantity and quality of input the child
(e.g., buying books and games) or personal (e.g., pro- receives than is nonverbal intelligence. Child-directed
viding attention or teaching). The confluence model
(Zajonc & Markus, 1975) considers siblings to be not
only consumers of the family’s resources but also part Corresponding Author:
of this pool of resources. According to this theory, the Naomi Havron, Laboratoire de Sciences Cognitives et Psycholinguistique,
Département d’Etudes Cognitives, École Normale Supérieure, École des
child’s intellectual environment is made up of the mean Hautes Etudes en Sciences Sociales, Centre Nationale de la Recherche
intellectual ability in the family; children, who have Scientifique, PSL University, 29 rue d’Ulm, 75005 Paris, France
lower intellectual abilities than adults, bring this mean E-mail: [email protected]
2 Havron et al.

speech from adults, especially in one-on-one interactions, have better language skills than children whose older
has been found to promote language learning (e.g., Hart siblings are less likely to do so, then this would suggest
& Risley, 1995; Ramírez-Esparza, García-Sierra, & Kuhl, that some older siblings do contribute to their younger
2014). The relationship between the number of older sib- siblings’ language development—in other words, that
lings and language development may relate to a decrease compensation exists.
in parents’ ability to spend one-on-one time with each Age gap might have an effect on parents’ ability to
child separately. For example, when both siblings are provide linguistic simulation. More narrowly spaced
present, parents may be more responsive to the older siblings might present greater competition for parental
child, at the expense of providing (quality) language input resources because of their own level of demands (as
to the younger child (Huttenlocher, Vasilyeva, Waterfall, they become older, they require less care and supervi-
Vevea, & Hedges, 2007; see also Hoff-Ginsberg, 1998). sion). Age gap might also have an effect on the likeli-
But can siblings themselves make up for some of the hood that the sibling can compensate for the loss of
lost parental resources by providing their own input parental linguistic simulation (Hoff-Ginsberg & Krueger,
and interactions? Though children may not be as adept 1991). The older the siblings are, the more likely they
as adults at adapting their speech to younger children are to have more developed social and linguistic skills,
(Mannle, Barton, & Tomasello, 1992), siblings are able thus providing better input. Indeed, one study found
to adjust their teaching strategies to their younger sib- that more narrowly spaced siblings had a larger nega-
lings’ ages (Dunn & Kendrick, 1982). Indeed, having tive effect on verbal test scores in high school than
older siblings positively affects children’s social- more widely spaced siblings did (Powell & Steelman
communicative skills (Hoff, 2006), the ability to join in Carr, 1990). As mentioned above, the confluence theory
conversations (Dunn & Shatz, 1989), and some aspects also makes this prediction. This leads us to predict that
of syntactic development, such as pronoun use (me– the larger the age gap between target child and sibling,
you; Oshima-Takane, Goodz, & Derevensky, 1996). In the less detrimental the effect of having a sibling.
addition, in research on bilingual families, older siblings A second variable of interest is gender. Older sisters
are found to be effective second-language models; their are more likely to engage in positive and nurturing
second-language skills can be better than the parents’, behavior than older brothers are (e.g., Tucker, McHale,
and they tend to use the second language with their & Crouter, 2001). Additionally, at early ages, girls tend
younger siblings (e.g., Bridges & Hoff, 2014; Duncan, to have more advanced language skills than boys (e.g.,
2017). This suggests that input and interactions with Eriksson et al., 2012). This advantage seems to last until
siblings can be beneficial, at least when input from the 5 to 6 years of age (Peyre et al., 2019), though it is
parents is of lesser quality. Conversely, other studies possible that a small advantage in language-related
have found that input from siblings does not contribute tasks remains throughout life (e.g., Ullman, Miranda, &
to vocabulary development of the target child and gen- Travers, 2008). Consequently, sisters might contribute
erally accounts for a very small percentage of input to better-quality input than brothers. All this leads to the
children in some cultures (Shneidman, Arroyo, Levine, prediction that having an older sister is better for one’s
& Goldin-Meadow, 2013). How can these discrepancies language development than having an older brother
be resolved? (i.e., an effect of the older sibling’s sex). Additionally,
We suggest that there are two main routes by which when an older sister is slightly older than the target
siblings may affect language development. By compet- child, she is relatively young and likely not very good
ing for parents’ attention, older siblings might be at providing compensatory input. In contrast, much
depriving their younger sibling of personally tailored older sisters should be better able to provide such
child-directed input. By providing their own input, stimulation; they also might be more likely to be del-
however, siblings may partly make up for some lost egated some of the caretaking of the target child and
stimulation. Finding a negative effect of older siblings thus be in a position to provide that stimulation. We
on language development could mean that, on average, therefore predicted that the age-gap effect would be
the effect of competition is larger than the effect of greater for older sisters than for older brothers.
compensation, or it could mean that older siblings are To test these predictions, we reanalyzed data from the
unable to compensate at all. EDEN cohort, a large population-based sample of French
To tease apart these two scenarios, we looked into children. Among children who have only one older sib-
two different characteristics of older siblings—namely, ling, we predicted better language scores for those who
age gap and sex. As we describe below, there is reason have an older sister than an older brother. We predicted
to believe that older siblings’ characteristics affect their worse language outcomes for children who were closer
ability and inclination to provide their younger siblings in age to their preceding sibling. We also predicted a
with their own quality input. If children whose older steeper slope of language scores on age gap when the
siblings are more likely to compensate for lost input older sibling was a sister (i.e., a larger difference between
The Effect of Older Siblings on Language Development 3

the language scores of children with an older sister and order in previous studies, e.g., Kristensen & Bjerkedal,
the language scores of children with an older brother 2007) with 80% power.
when the age gap was large than when it was small).
Materials
Method Predictors of language skills. Gestational age and
Study design birth weight were collected from obstetrical records, and
the older sibling’s sex and age were reported by the
We used data from the EDEN mother-child cohort study mother at the birth of the child included in the EDEN
(Heude et al., 2016), the primary aim of which is to cohort. Smoking status and alcohol consumption during
identify prenatal and early postnatal nutritional, envi- pregnancy (units per week) were determined from the
ronmental, and social determinants of children’s health questionnaires filled out by the mothers during preg-
and development. This is a longitudinal study, tracking nancy and at delivery. Mothers completed questionnaires
children’s development from before birth to the age of on partial or exclusive breastfeeding (breastfeeding initi-
11 years (so far). Participants were recruited between ation; Bernard et al., 2017). Both parents completed
2003 and 2006 from two university hospitals’ maternity questionnaires on their age at the child’s birth, family
units, both in France (in Poitiers and Nancy). Exclusion income, and education level. For level of parental educa-
criteria included history of diabetes, twin pregnancies, tion at birth of the child and household income (in thou-
intention to deliver outside the university hospital, sands of euros per month) at each age of testing, the
intention to move out of the study region within the averages for both parents were used in the analyses (see
next 3 years, and inability to speak French. The study the Statistical Analysis section). At each follow-up visit,
was approved by the Bicêtre Hospital Ethical Research parents completed questionnaires providing information
Committee and by the French data-protection authority on the sex and age of children born into the family after
(Commission Nationale de l’Informatique et des Libertés). the study started.
Written informed consent was obtained from parents—for
themselves at the time of enrollment and for the newborn Language skills (outcome variables). All tests and
after delivery. For the current study, we used language questionnaires described below were combined to create
measures taken at 2, 3, and 5 to 6 years of age (see Table a single score at each age, except at age 2, when only one
1 for details). The EDEN cohort study also contains many test was conducted. Using the same data set, Peyre and col-
other cognitive measures that are nonlinguistic and were leagues (2016) found that a single latent factor provided an
thus not part of the current analyses. excellent fit to the data at both 3 years and 5 to 6 years of
age, thus providing a general index of language skills.
Participants
Statistical analysis
There were 1,276 eligible children with language-skills
data available at 2, 3, or 5 to 6 years. 1 There were 1,154 All analyses were preregistered on the Open Science
children (483 with one older sibling) with language- Framework before they were performed (https://osf.io/
skills data available at 2 years; for some children (n = pgtyx/). They were performed as preregistered, except
122), language-skills data were available at 3 or 5 to 6 that in the adjusted models (introduced below), we
years but not at 2 years. There were 996 children (416 added a control factor for the effect of having a younger
with one older sibling) with language-skills data avail- sibling (results without controlling for this effect were
able at 3 years, and 898 children (381 with one older similar; see the Supplemental Material available online).
sibling) with language-skills data available at 5 to 6 Additional exploratory analyses can be found in the
years. About 46% of the older siblings were girls, and Supplemental Material as well.
6.88% of families were single-parent families. Language scores representing language skills at 3
Although it is difficult to accurately estimate statisti- years and at 5 to 6 years were calculated as the mean
cal power, given that the effect size of interest was of the scores at each time point (each score was first
unknown, that we had multiple hypotheses, and that converted into a z score in order for each test to have
we had unequal sample sizes at the different ages, we the same weight).
estimated that this sample size was large enough to There were a few missing data points on language
allow us to detect effects similar to those found in prior tests at 3 years of age and at 5 to 6 years of age (age
work. For example, with a minimum of 381 participants 3: 5.9% in our total sample and 5.3% in the subsample
per group (the smallest sample in any cell), we could of children with an older sibling only; ages 5–6: 1.9%
detect an effect size (d) of 0.2 (the effect size of birth in our total sample and 1.8% of children with an older
4 Havron et al.

Table 1. Language Tests Used to Assess Children at the Three Time Points in the Present Study

Age group Type of measure Test

Two-year-olds Parental French version of the CDI-2. Parents were asked to indicate which words from a
(M = 24 months, questionnaire list of 100 their child could say spontaneously (expressive vocabulary). The score
SD = 1) is the sum of the words produced by the child. The CDI-2 has high test-retest
reliability and strong associations with the corresponding scores from the longer
version (Kern et al., 2010).
Three-year-olds Assessment Items from the ELOLA battery, a European oral-language test battery:
(M = 38 months, by trained • Semantic Fluency, scored as the sum of the number of animals retrieved
SD = 1) psychologists spontaneously in 1 min plus the number of objects named in 1 min
• Word and Nonword Repetition, scored as the number of words (6 items) and
nonwords (6 items) repeated correctly
• Picture Naming, scored as the number of pictures named correctly (10 items,
e.g., “cheval” [horse])
Items from the NEPSY battery, a developmental neuropsychological assessment:
• Sentence Repetition, scored as the number of sentences of increasing complexity
and length repeated correctly (17 items, e.g., “dors bien” [sleep well])
Comprehension of Instructions, a sentence-comprehension task scored as the
number of correct answers by pointing at one of eight pictures (13 items, e.g.,
“montre moi un grand lapin” [show me a large rabbit])
Five- to six-year- Assessment Tests from the NEPSY battery:
olds (M = 68 by trained • Nonword Repetition, scored as the number of syllables repeated correctly (out
months, SD = 2) psychologists of 46 syllables in 13 nonwords—e.g., “kiutsa,” a nonword with two syllables)
• Sentence Repetition
Items from the WPPSI-III battery, a developmental neuropsychological assessment:
• Information, scored as the number of correct answers (speaking or pointing) to
questions that address a broad range of general-knowledge topics (34 items)
• Vocabulary, scored as the number of words correctly defined (25 items)
• Word Reasoning, scored as the number of concepts correctly identified from a
series of clues (28 items)

Note: CDI-2 = MacArthur-Bates Communicative Development Inventory (Kern, Langue, Zesiger, & Bovet, 2010); ELOLA = Batterie d’Évaluation du
langage oral de l’enfant aphasique (De Agostini et al., 1998); NEPSY = NEPSY Bilan Neuropsychologique de l’Enfant (Kemp, Kirk, & Korkman,
2001; Korkman, Kirk, & Kemp, 2003); WPPSI-III = Wechsler Preschool and Primary Scale of Intelligence–3rd edition (Wechsler, 1967).

sibling only). For predictors of language skills, fewer repeated measure, and the presence of one older sib-
than 1% of the data points were missing in both analysis ling as the independent variable (participant was
samples. Missing data on predictors of language skills entered as a random effect). A second model was
were determined using multiple imputation (n = 50), adjusted for exact age at time of evaluation (continu-
assuming that missing data were probably not missing ous) and other predictors of language skills—sex, ges-
at random (Donders, van der Heijden, Stijnen, & Moons, tational age (weeks), birth weight (kg), maternal age at
2006). All analyses were performed using SAS software delivery (years), paternal age at delivery (years), breast-
(Version 9.4). Multiple imputations were implemented feeding initiation (%), alcohol use during pregnancy
using the SAS PROC MI procedure with the fully con- (units per week), tobacco use during pregnancy (%),
ditional specification statement. parental education (years), household income (in thou-
sands of euros per month), and presence of younger
Analysis 1: having a sibling versus having no sib- siblings at the time of evaluation. These are the main
ling. Before looking into the mediating role of sex and factors suspected of influencing cognitive development
age gap, we checked that having an older sibling (either that were available in the EDEN cohort study (Peyre,
a brother or a sister) was indeed detrimental for language Galera, et al., 2016).
outcomes, as found in previous research (e.g., Black et al., We adjusted the model for these control variables by
2005) and by Peyre, Bernard, et al. (2016), when analyz- fitting three linear models, one for each testing age, with
ing 5- to 6-year-olds’ data in the same cohort. the child’s precise age at testing as well as the control
In our sample (N = 1,276), we ran a linear mixed variables (but not with our hypothesized predictors).
regression model with language skills as the dependent We then extracted the residuals from these fitted models
variable, testing age (2, 3, and 5–6 years of age) as a and combined them into one data set, which served as
The Effect of Older Siblings on Language Development 5

the outcome for the mixed-effects regression analysis association was not significant in the model adjusted
mentioned above. The reason the adjustment was done for the other predictors of language skills (d = −0.04,
this way was to enable us to control for exact age at p = .071). No significant interaction between the sex of
time of evaluation, which would not be possible if the the older sibling and age gap was found (see Fig. 2).
adjustments were done directly within the main model. Finally, in the follow-up analyses, children with an
older sister had similar language skills to children with-
Analysis 2: sex, age gap, and interaction. Next, we out an older sibling, while children with an older
tested our prediction of a positive effect of older sisters brother had significantly worse language skills (see
on language development (compared with older broth- Table 5 and Fig. 3).
ers), a positive effect of a larger age gap, and a stronger
age-gap effect for older sisters.
Discussion
In the subsample of children with only one older
sibling (n = 547), we performed a linear mixed-effects Having an older sibling is associated with lower verbal
regression model with language skills as the dependent skills compared with having no older sibling (e.g.,
variable, testing age as a repeated measure, sex of the Black et al., 2005). We reanalyzed data from the EDEN
older sibling and age gap (i.e., age difference between cohort to examine the effect of the sex of the older
the two siblings) as independent variables, and the sibling and the age gap between siblings on the younger
interaction between the older sibling’s sex and the age child’s language development, reasoning that some
gap. Participant was entered as a random effect. In a older siblings are more likely to compensate for paren-
second model, we adjusted for exact age at time of tal resources than others.
evaluation and the other predictors of language skills, Our hypothesis that children with older sisters would
as described above. have higher language scores than children with older
brothers was confirmed. The size of the effect of sibling
Preregistered follow-up analyses. We compared chil- sex, d = 0.22, although small, is not negligible in epi-
dren who had no sibling with children who had either an demiological studies of cognitive development. It would
older sister or an older brother (unlike in Analysis 1, in be the equivalent of 3 IQ points and is similar to the
which the gender of the older sibling was not specified). birth-order effect in Kristensen and Bjerkedal’s (2007)
We did this in order to estimate the effects of older broth- seminal study. In a follow-up comparison separating
ers and sisters, respectively, relative to the baseline of having siblings by sex, we found that although children with
no older sibling. We ran a linear mixed-effects regression an older brother exhibited lower language skills than
model with language skills as the dependent variable and children with no older sibling, children who had an
with testing age and presence and sex of the older sibling as older sister scored comparably with children with no
independent variables (categorical with three levels—no older sibling. It thus might be more accurate to think
sibling, brother, and sister—and with no sibling as the refer- of the well-established negative older-sibling effect as
ence). Participant was entered as a random effect. A second an older-brother effect.
model was adjusted for exact age at time of evaluation and It is unclear whether older sisters help language
the other predictors of language skills, as described above. development—or just hinder it less than older brothers
do. Indeed, there are two possible reasons for this
effect. First, as argued in the introduction, older sisters
Results may themselves contribute to their younger sibling’s
Characteristics of participants in the samples of analysis language development. Older sisters may be more pre-
are presented in Table 2. As predicted, Analysis 1 disposed or better trained to act as caregivers than older
showed that language skills of children with one older brothers (e.g., Tucker et al., 2001)—or even simply
sibling were lower than those of children without an more talkative or more willing playmates. Young girls
older sibling (Cohen’s d = −0.14 and −0.17 in the unad- are also more linguistically advanced than young boys
justed and adjusted models, respectively; see Table 3 until at least age 5 to 6 (Peyre et al., 2019), and thus
and Fig. 1). Also as predicted, Analysis 2 showed that they may be better able to provide quality input. An
language skills of children with an older sister were alternative explanation suggested by our results is that
higher than those of children with an older brother an older brother may be more demanding for parents
(d = 0.26 and 0.22 in the unadjusted and adjusted mod- than an older sister, at the expense of the younger
els, respectively; see Table 4 and Figs. 2 and 3). Con- sibling. Mothers of baby boys experience more stress
trary to our prediction, results of Analysis 2 showed than mothers of baby girls (Scher & Sharabany, 2010),
that language skills were negatively associated with age and later on, boys show more externalizing behavior
gap in the unadjusted model (d = −0.05, p = .003). This and tend to be more irritable (Leaper, 2002). This could
6 Havron et al.

Table 2. Characteristics of Children in the Analysis Sample With No Older Sibling or One Older Sibling

Children with no sibling Children without an Children with one

or one older sibling older sibling older sibling
Variable (n = 1,276) (n = 729) (n = 547)
Children with one older sibling 43.3% 0% 100%
Sex of the older sibling (male) — — 53.9%
Age gap (years) — — M = 3.7 (SD = 2.2)
Language skills at 2 years n = 1,154 n = 671 n = 483
CDI-2 score M = 62.1 (SD = 29.2) M = 64.5 (SD = 28.6) M = 58.7 (SD = 29.6)
Age of the child at the time of CDI-2 (months) M = 24.3 (SD = 1.1) M = 24.4 (SD = 1.8) M = 24.3 (SD = 0.7)
Language skills at 3 years n = 996 n = 580 n = 416
Semantic Fluencya M = 0.0 (SD = 0.8) M = 0.1 (SD = 0.8) M = 0.0 (SD = 0.8)
Word and Nonword Repetitiona M = 0.0 (SD = 0.9) M = 0.1 (SD = 0.9) M = 0.0 (SD = 1.0)
Sentence Repetition M = 7.2 (SD = 3.3) M = 7.2 (SD = 3.1) M = 7.2 (SD = 3.5)
Picture Naming M = 7.0 (SD = 1.8) M = 7.2 (SD = 1.8) M = 6.9 (SD = 1.9)
Comprehension of Instructions M = 8.6 (SD = 2.9) M = 8.8 (SD = 2.9) M = 8.4 (SD = 3.0)
Age of the child at the time of tests (in months) M = 38.0 (SD = 0.8) M = 38.0 (SD = 0.8) M = 38.0 (SD = 0.8)
Language skills at 5–6 years n = 898 n = 517 n = 381
Nonword Repetition M = 21.0 (SD = 4.9) M = 21.0 (SD = 4.9) M = 21.0 (SD = 5.0)
Sentence Repetition M = 15.6 (SD = 4.0) M = 15.8 (SD = 3.9) M = 15.2 (SD = 4.1)
Information M = 25.1 (SD = 2.9) M = 25.4 (SD = 2.9) M = 24.7 (SD = 2.8)
Vocabulary M = 23.8 (SD = 5.6) M = 24.3 (SD = 5.6) M = 23.2 (SD = 5.6)
Word Reasoning M = 16.3 (SD = 4.7) M = 16.6 (SD = 4.6) M = 15.8 (SD = 4.8)
Age of the child at the time of tests (in months) M = 67.9 (SD = 1.8) M = 67.9 (SD = 1.7) M = 68.0 (SD = 1.9)
Predictor of cognitive skills
Sex (male) 51.7% 51.5% 52.1%
Gestational age (in weeks) M = 39.2 (SD = 1.7) M = 39.3 (SD = 1.8) M = 39.2 (SD = 1.6)
Birth weight (kg) M = 3.26 (SD = 0.51) M = 3.22 (SD = 0.51) M = 3.30 (SD = 0.50)
Mother’s age at delivery (years) M = 28.5 (SD = 4.5) M = 27.3 (SD = 4.4) M = 29.9 (SD = 4.3)
Father’s age at delivery (years) M = 31.3 (SD = 5.6) M = 30.1 (SD = 5.5) M = 32.9 (SD = 5.2)
Breastfeeding initiation 73.1% 75.6% 69.8%
Alcohol during pregnancy (drinks per week) M = 0.56 (SD = 1.60) M = 0.52 (SD = 1.61) M = 0.62 (SD = 1.61)
Maternal smoking during pregnancy 24.8% 26.3% 23.0%
Parental education (years) M = 13.6 (SD = 2.3) M = 13.6 (SD = 2.3) M = 13.6 (SD = 2.4)
Household income at 2 years (k€) M = 2.61 (SD = 1.01) M = 2.53 (SD = 1.00) M = 2.71 (SD = 1.01)
Younger siblings at 2 years 8.4% 13.1% 2.6%
Younger siblings at 3 years 20.1% 28.4% 9.5%
Younger siblings at 5–6 years 35.2% 47.1% 19.7%
Center (Nancy, France) 50.1% 46.9% 54.3%

Note: CDI-2 = MacArthur-Bates Communicative Development Inventory (Kern, Langue, Zesiger, & Bovet, 2010). aScores have been standardized
in the whole data set.

mean that there is less competition for parental language scores. This hypothesis was not supported. If
resources from an older sister than from an older anything, we found a trend in the opposite direction:
brother. Although our results cannot help us discrimi- The more closely spaced the siblings were, the higher
nate between these two explanations, both focus on the language scores of the target child. Thus, although
the linguistic environment available to children and the the present study was well powered, the age-gap effect
types of linguistic experiences and input they are remains uncertain. An even larger study would be nec-
exposed to in their early years of life. Our results thus essary to determine whether or not there is a genuine
join a vast literature on the effect of the quantity and negative age-gap effect. This result fails to support the
quality of input on language development (e.g., confluence model discussed in the introduction, which
Ramírez-Esparza et al., 2014). states that the older the siblings, the higher their intel-
We additionally hypothesized that the age gap lectual abilities and the less disruptive they are to their
between siblings would be positively correlated with younger siblings’ development (Zajonc & Markus, 1975).
The Effect of Older Siblings on Language Development 7

2 A limitation of the current study is that demographic

Language Skills (standardized)
characteristics of the older siblings were used as a
1 proxy for the level of stimulation the older sibling
would provide the younger sibling, as well as for the
0 level of competition. We have no direct evidence that
the interactions between children and their older sisters
–1 were different from children’s interactions with their
older brothers, or that the interactions between the
–2 caregiver and the key child differed in the presence of
an older brother compared with an older sister. More-
–3 over, our sample drew from the French population, and
the results may not generalize to other cultures. Indeed,
No Older One Older Frank, Braginsky, Marchman, and Yurovsky (2019; see
Sibling Sibling also Steelman, Powell, Werum, & Carter, 2002) found
large variability in the birth-order effects across differ-
Fig. 1. Language skills of children with no older siblings com-
pared with those who had one older sibling. The boxes encompass
ent countries, compared with, for example, stability in
interquartile ranges, the black stars represents means, and the error the female advantage in early language development.
bars represent ±2 SD. Open circles represent individual data points. Another limitation is that our sample of children with
Because the model included children’s scores at each age separately, and without an older sibling differed in their demo-
children have between one and three data points apiece, depending
on whether data were missing. graphic characteristics (e.g., children with no older

Table 3. Results of Linear Mixed Regression Models With Language Skills as the Dependent
Variable and the Presence of an Older Sibling as the Independent Variable (N = 1,276)

Model and predictor β SD p

Unadjusted model: older sibling (reference: without an older sibling) −0.141 0.030 < .0001
Model 1 (adjusted): older sibling (reference: without an older sibling) −0.168 0.037 < .0001

Note: For this analysis, language scores were available for 1,154 two-year-olds (483 of whom had an older sibling),
996 three-year-olds (416 of whom had an older sibling), and 898 five- to six-year-olds (381 of whom had an
older sibling). Adjusted Model 1 was adjusted for exact age at time of evaluation and other predictors of cognitive
development—sex, gestational age (weeks), birth weight (kg), maternal age at delivery (years), paternal age at
delivery (years), breastfeeding initiation (%), alcohol use during pregnancy (units per week), tobacco use during
pregnancy (%), parental education (years), household income (thousands of euros per month), and presence of
younger siblings at the time of evaluation.

Table 4. Results of Linear Mixed Regression Models With Language Skills

as the Dependent Variable and Sex of the Older Sibling, Age Gap Between
the Two Siblings, and the Interaction Between Sex and Age Gap as the
Independent Variables (N = 547)

Model and predictor β SD p

Unadjusted model
Sex of older sibling (male; reference: female) −0.259 0.091 .0045
Age gap (years) −0.048 0.016 .0031
Age Gap × Sex of Older Sibling 0.020 0.021 .3543
Model 1 (adjusted)
Sex of older sibling (male; reference: female) −0.218 0.108 .0433
Age gap (years) −0.035 0.019 .0708
Age Gap × Sex of Older Sibling 0.008 0.025 .7411

Note: For this analysis, language scores were available for 483 two-year-olds (222 of
whom had an older sister), 416 three-year-olds (187 of whom had an older sister),
and 381 five- to six-year-olds (177 of whom had an older sister). Adjusted Model 1
was adjusted for exact age at time of evaluation and other predictors of cognitive
development—sex, gestational age (weeks), birth weight (kg), maternal age at delivery
(years), paternal age at delivery (years), breastfeeding initiation (%), alcohol use during
pregnancy (units per week), tobacco use during pregnancy (%), parental education
(years), household income (thousands of euros per month), and presence of younger
siblings at the time of evaluation.
8 Havron et al.

3
Older Brother
2 Older Sister

Language Skills (standardized) 1

–1

–2

–3
0 50 100 150
Age Gap Between the Two Siblings (months)
Fig. 2. Scatterplot showing the relationship between language skills and age gap between par-
ticipants and their siblings, separately for children with an older brother and an older sister. Solid
lines indicate best-fitting regressions, and dotted lines represent 95% confidence intervals.

sibling tended to also have more younger siblings). To conclude, the current study found that having an
Although we adjusted for these factors statistically, it older sister was associated with higher language scores
would be helpful for future studies to use a more bal- than having an older brother. Children who had an
anced sample. older sister scored similarly to children with no older

2
Language Skills (standardized)

–1

–2

–3

Older No Older Older

Brother Sibling Sister
Fig. 3. Language skills of children with no older siblings compared with those who
had one older brother or one older sister. The boxes encompass within-quartile ranges,
the black stars represents means, and the error bars represent ±2 SD. Open circles
represent individual data points. Because the model included children’s scores at each
age separately, children have between one and three data points apiece, depending
on whether data were missing.
The Effect of Older Siblings on Language Development 9

Table 5. Results of Linear Mixed Regression Models With Language Skills as

the Dependent Variable and the Presence and Sex of the Older Sibling as the
Independent Variables (N = 1,276)

Model and predictor β SD p

Unadjusted model
Older brother (reference: without an older sibling) −0.234 0.037 < .0001
Older sister (reference: without an older sibling) −0.042 0.040 .2969
Model 1 (adjusted)
Older brother (reference: without an older sibling) −0.267 0.045 < .0001
Older sister (reference: without an older sibling) −0.060 0.047 .2085

Note: For this analysis, language scores were available for 1,154 two-year-olds (483 of whom
had an older sibling), 996 three-year-olds (416 of whom had an older sibling), and 898 five- to
six-year-olds (381 of whom had an older sibling). Adjusted Model 1 was adjusted for exact
age at time of evaluation and other predictors of cognitive development—sex, gestational age
(weeks), birth weight (kg), maternal age at delivery (years), paternal age at delivery (years),
breastfeeding initiation (%), alcohol use during pregnancy (units per week), tobacco use during
pregnancy (%), parental education (years), household income (thousands of euros per month),
and presence of younger siblings at the time of evaluation.

sibling, whereas children who had an older brother Declaration of Conflicting Interests
scored significantly lower. We found no evidence that The author(s) declared that there were no conflicts of interest
a larger age gap between siblings was beneficial for the with respect to the authorship or the publication of this
language development of the younger child. article.

Action Editor Funding

Rebecca Treiman served as action editor for this article. The EDEN study was supported by the Foundation for Medi-
cal Research (FRM); National Agency for Research (ANR);
Author Contributions National Institute for Research in Public Health (IRESP: Très
Grandes Infrastructures de Recherche, or TGIR, cohorte santé
N. Havron developed the study concept and wrote the first
2008 program); French Ministry of Health (DGS); French Min-
draft of the manuscript. F. Ramus, A. Cristia, H. Peyre, B.
istry of Research; INSERM Bone and Joint Diseases National
Heude, and A. Forhan commented on drafts of the manu-
Research (PRO-A) and Human Nutrition National Research
script. B. Heude and A. Forhan collected the original data on
programs; Paris-Sud University; Nestlé; French National Insti-
which the study was based. N. Havron, F. Ramus, A. Cristia,
tute for Population Health Surveillance (InVS); French
and H. Peyre constructed the data-analysis plan; H. Peyre
National Institute for Health Education (INPES); the European
analyzed the data. All authors approved the final version of
Union Seventh Framework Programme for Research and Tech-
the manuscript for submission.
nological Development (FP7/2007-2013; High-End Climate
Impacts and Extremes, or HELIX; European Study of Cohorts
ORCID iD for Air Pollution Effects, or ESCAPE; Environmental Health
Naomi Havron https://orcid.org/0000-0001-6429-1546 Risks In European Birth Cohorts, or ENRIECO; and Mecha-
nisms of the Development of Allergy, or MEDALL, projects);
Acknowledgments Diabetes National Research Program (through a collaboration
with the French Association of Diabetic Patients, or AFD);
We are grateful to the participating families, the midwife research
French Agency for Food, Environmental and Occupational
assistants (L. Douhaud, S. Bedel, B. Lortholary, S. Gabriel,
Health & Safety (ANSES; formerly the French Agency for
M. Rogeon, and M. Malinbaum) for data collection, the psycholo-
Environmental Health Safety); Mutuelle Générale de
gists (Marie-Claire Cona and Marielle Paquinet), and P. Lavoine,
l’Education Nationale (MGEN); and the French Speaking Asso-
J. Sahuquillo, and G. Debotte for checking, coding, and data
ciation for the Study of Diabetes and Metabolism (ALFE-
entry. We also acknowledge the commitment of the members
DIAM). Additional funding came from ANR contracts
of the EDEN Mother-Child Cohort Study Group: I. Annesi-Maesano,
ANR-17-EURE-0017, ANR-11-0001-02 PSL, and ANR-12-DSSA-
J. Y. Bernard, J. Botton, M. A. Charles, P. Dargent-Molina,
0005-01.
B. de Lauzon-Guillain, P. Ducimetière, M. De Agostini, B. Foliguet,
A. Forhan, X. Fritel, A. Germa, V. Goua, R. Hankard, B. Heude,
M. Kaminski, B. Larroque, N. Lelong, J. Lepeule, G. Magnin, Supplemental Material
L. Marchand, C. Nabet, F. Pierre, R. Slama, M. J. Saurel-Cubizolles, Additional supporting information can be found at http://
M. Schweitzer, and O. Thiebaugeorges. journals.sagepub.com/doi/suppl/10.1177/0956797619861436
10 Havron et al.

Open Practices Duncan, T. G. S. (2017). How does maternal education influ-

ence language acquisition? Interdependencies between
TC
environment and input in the bilingual development of
immigrant and refugee children. Unpublished doctoral
The design and analysis plans for the experiments were pre- dissertation, University of Alberta, Canada.
registered at https://osf.io/pgtyx/. Changes made to the pre- Dunn, J., & Kendrick, C. (1982). The speech of two- and
registration after data collection are described in the main three-year-olds to infant siblings: ‘Baby talk’ and the con-
text. The data used in this study can be accessed through the text of communication. Journal of Child Language, 9,
EDEN Mother-Child Cohort Study Group website (http://eden 579–595. doi:10.1017/S030500090000492X
.vjf.inserm.fr/index.php/fr/) after completion of a request or Dunn, J., & Shatz, M. (1989). Becoming a conversationalist
registration form. The complete Open Practices Disclosure for despite (or because of) having an older sibling. Child
this article can be found at http://journals.sagepub.com/doi/ Development, 60, 399–410. doi:10.2307/1130985
suppl/10.1177/0956797619861436. This article has received the Eriksson, M., Marschik, P. B., Tulviste, T., Almgren, M.,
badge for Preregistration. More information about the Open Pérez Pereira, M., Wehberg, S., . . . Gallego, C. (2012).
Practices badges can be found at http://www.psychological Differences between girls and boys in emerging lan-
science.org/publications/badges. guage skills: Evidence from 10 language communities.
British Journal of Developmental Psychology, 30, 326–343.
doi:10.1111/j.2044-835X.2011.02042.x
Note
Frank, M. C., Braginsky, M., Marchman, V., & Yurovsky, D.
1. We analyzed a subsample of the EDEN cohort full sample (2019). Variability and consistency in early language
consisting only of children with one older sibling or no older learning: The Wordbank project. Retrieved from https://
siblings. Children with more than one older sibling (n = 354) langcog.github.io/wordbank-book/index.html
were excluded because additional older siblings may obscure Hart, B., & Risley, T. R. (1995). Meaningful differences in
the effects of age gap and sex of the immediately preceding the everyday experience of young American children.
older sibling. The EDEN cohort full sample included 1,907 live- Baltimore, MD: Paul H. Brookes.
born children, as described in detail by Heude et al. (2016). The Heude, B., Forhan, A., Slama, R., Douhaud, L., Bedel, S.,
attrition rate of children at age 5 years was 41%, both for the Saurel-Cubizolles, M.-J., . . . Thiebaugeorges, O. (2016).
overall sample in the cohort and for our analysis sample. Cohort profile: The EDEN mother-child cohort on the
prenatal and early postnatal determinants of child health
and development. International Journal of Epidemiology,
References 45, 353–363. doi:10.1093/ije/dyv151
Bernard, J. Y., Armand, M., Peyre, H., Garcia, C., Forhan, A., Hoff, E. (2006). How social contexts support and shape lan-
& De Agostini, M., . . . the EDEN Mother-Child Cohort guage development. Developmental Review, 26, 55–88.
Study Group. (2017). Breastfeeding, polyunsaturated fatty doi:10.1016/j.dr.2005.11.002
acid levels in colostrum and child intelligence quotient at Hoff-Ginsberg, E. (1998). The relation of birth order and
age 5-6 years. The Journal of Pediatrics, 183, 43–50.e3. socioeconomic status to children’s language experience
doi:10.1016/j.jpeds.2016.12.039 and language development. Applied Psycholinguistics, 19,
Black, S. E., Devereux, P. J., & Salvanes, K. G. (2005). The 603–629. doi:10.1017/s0142716400010389
more the merrier? The effect of family size and birth Hoff-Ginsberg, E., & Krueger, W. (1991). Older siblings as
order on children’s education. The Quarterly Journal of conversational partners. Merrill-Palmer Quarterly: Journal
Economics, 120, 669–700. doi:10.1093/qje/120.2.669 of Developmental Psychology, 37, 465–481.
Blake, J. (1981). Family size and the quality of children. Huttenlocher, J., Vasilyeva, M., Waterfall, H. R., Vevea, J. L.,
Demography, 18, 421–442. doi:10.2307/2060941 & Hedges, L. V. (2007). The varieties of speech to young
Bridges, K., & Hoff, E. (2014). Older sibling influences on children. Developmental Psychology, 43, 1062–1083.
the language environment and language development of doi:10.1037/0012-1649.43.5.1062
toddlers in bilingual homes. Applied Psycholinguistics, 35, Kemp, S. L., Kirk, U., & Korkman, M. (2001). Essentials of
225–241. doi:10.1017/S0142716412000379 NEPSY Assessment (1st ed.). Hoboken, NJ: Wiley.
De Agostini, M., Metz-Lutz, M. N., Van Hout, A., Chavance, M., Kern, S., Langue, J., Zesiger, P., & Bovet, F. (2010). Adaptations
Deloche, G., Pavao-Martins, I., & Dellatolas, G. (1998). françaises des versions courtes des inventaires du dével-
Batterie d’évaluation du langage oral de l’enfant apha- oppement communicatif de MacArthur-Bates [French adap-
sique (ELOLA): Standardisation française (4–12 ans) [Oral tations of short versions of MacArthur-Bates Communicative
language evaluation battery of aphasic children: A French Development Inventories]. ANAE: Approche neuropsy-
standardization]. Revue de neuropsychologie, 8(3), 319–367. chologique des apprentissages chez l’enfant, 107/108,
Donders, A. R. T., van der Heijden, G. J. M. G., Stijnen, T., 217–228.
& Moons, K. G. M. (2006). Review: A gentle introduc- Korkman, M., Kirk, U., & Kemp, S. (2003). Nepsy Bilan Neuro­
tion to imputation of missing values. Journal of Clinical psychologique de l’enfant [NEPSY Neuropsychological
Epidemiology, 59, 1087–1091. doi:10.1016/j.jclinepi.2006 As­sessment of Children]. Paris, France: Editions du Centre
.01.014 de Psychologie Appliquée.
The Effect of Older Siblings on Language Development 11

Kristensen, P., & Bjerkedal, T. (2007). Explaining the relation Powell, B., & Steelman Carr, L. (1990). Beyond sibship size:
between birth order and intelligence. Science, 316, 1717. Sibling density, sex composition, and educational out-
doi:10.1126/science.1141493 comes. Social Forces, 69, 181–206. doi:10.1093/sf/69.1.181
Leaper, C. (2002). Parenting boys and girls. In M. H. Bornstein Ramírez-Esparza, N., García-Sierra, A., & Kuhl, P. K. (2014).
(Ed.), Handbook of parenting, Vol. 1: Children and par- Look who’s talking: Speech style and social context in
enting (pp. 189–225). Hillsdale, NJ: Erlbaum. language input to infants are linked to concurrent and
Mannle, S., Barton, M., & Tomasello, M. (1992). Two-year- future speech development. Developmental Science, 17,
olds’ conversations with their mothers and preschool- 880–891. doi:10.1111/desc.12172
aged siblings. First Language, 12, 57–71. doi:10.1177/014 Scher, A., & Sharabany, R. (2010). Parenting anxiety and
272379201203404 stress: Does gender play a part at 3 months of age? The
Oshima-Takane, Y., Goodz, E., & Derevensky, J. L. (1996). Journal of Genetic Psychology, 166, 203–213. doi:10.3200/
Birth order effects on early language development: Do GNTP.166.2.203-214
secondborn children learn from overheard speech? Child Shneidman, L. A., Arroyo, M. E., Levine, S. C., & Goldin-
Development, 67, 621–634. doi:10.1111/j.1467-8624.1996. Meadow, S. (2013). What counts as effective input for
tb01755.x word learning? Journal of Child Language, 40, 672–686.
Peyre, H., Bernard, J. Y., Hoertel, N., Forhan, A., Charles, doi:10.1017/S0305000912000141
M.-A., & De Agostini, M., . . . the EDEN Mother-Child Steelman, L. C., Powell, B., Werum, R., & Carter, S. (2002).
Cohort Study Group. (2016). Differential effects of factors Reconsidering the effects of sibling configuration: Recent
influencing cognitive development at the age of 5-to-6 advances and challenges. Annual Review of Sociology,
years. Cognitive Development, 40, 152–162. doi:10.1016/j 28, 243–269. doi:10.1146/annurev.soc.28.111301.093304
.cogdev.2016.10.001 Tucker, C. J., McHale, S. M., & Crouter, A. C. (2001). Con­
Peyre, H., Galera, C., van der Waerden, J., Hoertel, N., ditions of sibling support in adolescence. Journal of Family
Bernard, J. Y., & Melchior, M., . . . the EDEN Mother-Child Psychology, 15, 254–271.
Cohort Study Group. (2016). Relationship between early Ullman, M. T., Miranda, R. A., & Travers, M. L. (2008).
language skills and the development of inattention/hyper- Sex differences in the neurocognition of language. In
activity symptoms during the preschool period: Results J. B. Becker, K. J. Berkley, N. Geary, E. Hampson, J. P.
of the EDEN mother-child cohort. BMC Psychiatry, 16, Herman, & E. A. Young (Eds.), Sex differences in the
Article 380. doi:10.1186/s12888-016-1091-3 brain from genes to behavior (pp. 291–309). New York,
Peyre, H., Hoertel, N., Bernard, J. Y., Rouffignac, C., Forhan, NY: Oxford University Press.
A., Taine M., . . . EDEN Mother–Child Cohort Study Group. Wechsler, D. (1967). Wechsler Preschool and Primary
(2019). Sex differences in psychomotor development dur- Scale of Intelligence–3rd edition. San Antonio, TX: The
ing the preschool period: A longitudinal study of the Psychological Corp.
effects of environmental factors and of emotional, behav- Zajonc, R. B., & Markus, G. B. (1975). Birth order and intel-
ioral, and social functioning. Journal of Experimental Child lectual development. Psychological Review, 82, 74–88.
Psychology, 178, 369–384. doi:10.1016/j.jecp.2018.09.002 doi:10.1037/h0076229