Nature and nurture in fussy eating from toddlerhood to early adolescence: findings from the Gemini twin cohort

Abstract

Background
Food fussiness (FF) describes the tendency to eat a small range of foods, due to pickiness and/or reluctance to try new foods. A common behaviour during childhood, and a considerable cause of caregiver concern; its causes are poorly understood. This is the first twin study of genetic and environmental contributions to the developmental trajectory of FF from toddlerhood to early adolescence, and stability and change over time.

Methods
Participants were from Gemini, a population‐based British cohort of n = 4,804 twins born in 2007. Parents reported on FF using the Child Eating Behaviour Questionnaire ‘FF’ scale when children were 16 months (n = 3,854), 3 (n = 2,666), 5 (n = 2,098), 7 (n = 703), and 13 years old (n = 970). A mixed linear model examined the trajectory of FF, and a correlated factors twin model quantified genetic and environmental contributions to variation in and covariation between trajectory parameters. A longitudinal Cholesky twin model examined genetic and environmental influences on FF at each discrete age.

Results
We modelled a single FF trajectory for all children, which was characterised by increases from 16 months to 7 years, followed by a slight decline from 7 to 13 years. All trajectory parameters were under strong genetic influence (>70%) that was largely shared, indicated by high genetic correlations. Discrete age analyses showed that genetic influence on FF increased significantly after toddlerhood (16 months: 60%, 95% CI: 53%–67%; 3 years: 83%; 81%–86%), with continuing genetic influence as indicated by significant genetic overlap across every age. Shared environmental influences were only significant during toddlerhood. Unique environmental influences explained 15%–26% of the variance over time, with some enduring influence from 5 years onwards.

Conclusions
Individual differences in FF were largely explained by genetic factors at all ages. Fussy eating also shows a significant proportion of environmental influence, especially in toddlerhood, and may, therefore, benefit from early interventions throughout childhood. Future work needs to refine the FF trajectory and explore specific trajectory classes.

Full text

Nature and nurture in fussy eating from toddlerhood
to early adolescence: findings from the Gemini twin
cohort
Zeynep Nas,
1
Moritz Herle,
2
Alice R. Kininmonth,
1,3
Andrea D. Smith,
4
Rachel Bryant-Waugh,
5
Alison Fildes,
6
and Clare H. Llewellyn
1
1
Department of Behavioural Science & Health, University College London, London, UK;
2
Social, Genetic &
Developmental Psychiatry Centre, Institute of Psychiatry, Psychology, and Neuroscience, King’s College London,
London, UK;
3
School of Food Science and Nutrition, University of Leeds, Leeds, UK;
4
MRC Epidemiology Unit,
University of Cambridge, Cambridge, UK;
5
South London and Maudsley NHS Foundation Trust, London, UK;
6
School
of Psychology, University of Leeds, Leeds, UK
Background: Food fussiness (FF) describes the tendency to eat a small range of foods, due to pickiness and/or
reluctance to try new foods. A common behaviour during childhood, and a considerable cause of caregiver concern;
its causes are poorly understood. This is the first twin study of genetic and environmental contributions to the
developmental trajectory of FF from toddlerhood to early adolescence, and stability and change over time. Methods:
Participants were from Gemini, a population-based British cohort of n=4,804 twins born in 2007. Parents reported
on FF using the Child Eating Behaviour Questionnaire ‘FF’ scale when children were 16 months (n=3,854), 3
(n=2,666), 5 (n=2,098), 7 (n=703), and 13 years old (n=970). A mixed linear model examined the trajectory of
FF, and a correlated factors twin model quantified genetic and environmental contributions to variation in and
covariation between trajectory parameters. A longitudinal Cholesky twin model examined genetic and environmental
influences on FF at each discrete age. Results: We modelled a single FF trajectory for all children, which was
characterised by increases from 16 months to 7 years, followed by a slight decline from 7 to 13 years. All trajectory
parameters were under strong genetic influence (>70%) that was largely shared, indicated by high genetic
correlations. Discrete age analyses showed that genetic influence on FF increased significantly after toddlerhood
(16 months: 60%, 95% CI: 53%–67%; 3 years: 83%; 81%–86%), with continuing genetic influence as indicated by
significant genetic overlap across every age. Shared environmental influences were only significant during
toddlerhood. Unique environmental influences explained 15%–26% of the variance over time, with some enduring
influence from 5 years onwards. Conclusions: Individual differences in FF were largely explained by genetic factors
at all ages. Fussy eating also shows a significant proportion of environmental influence, especially in toddlerhood,
and may, therefore, benefit from early interventions throughout childhood. Future work needs to refine the FF
trajectory and explore specific trajectory classes. Keywords: Longitudinal studies; twins; eating behaviour.
Introduction
Food fussiness describes the tendency to eat a limited
range of foods, often due to pickiness regarding
flavour or texture, and/or the reluctance to try new
foods and flavours. FF or picky eating is common and
typically develops early in life (during toddlerhood)
with prevalence rates ranging between 6% and 50%
(Machado, Dias, Lima, Campos, & Gonc
ßalves, 2016;
Mascola, Bryson, & Agras, 2010; Taylor, Wernimont,
Northstone, & Emmett, 2015). There have been few
studies of the developmental trajectory of fussy and
picky eating, which suggests the behaviours tend to
peak in early-to-middle childhood, with most, though
not all, children showing decreases in FF as they
mature into adolescence (Cardona Cano et al., 2015;
de Barse et al., 2015; Herle et al., 2020; Taylor &
Emmett, 2019; Taylor, Steer, Hays, & Emmett, 2019).
Persistent and/or severe fussiness can contribute to
detrimental physical and psychological health out-
comes, such as nutritional deficiencies, weight falter-
ing, and anxiety over food (Lafraire, Rioux, Giboreau,
& Picard, 2016; Taylor et al., 2015). Severe FF causes
considerable caregiver anxiety, can disrupt family
mealtimes (Wolstenholme, Heary, & Kelly, 2019), and
challenges family dynamics (Zucker et al., 2015).
Some epidemiological research has also linked FF in
childhood with increased risk for disordered eating in
adolescence and young adulthood (Carter Leno,
Micali, Bryant-Waugh, & Herle, 2022; Herle
et al., 2020; McClelland, Robinson, Potterton, Mount-
ford, & Schmidt, 2020). Excessive selective eating can
be a key symptom of avoidant/restrictive food intake
disorder (ARFID), a relatively recently recognised
eating disorder included in the DSM-5 (Zimmerman
& Fisher, 2017). In summary, FF is common, poten-
tially important to the healthy development of chil-
dren, and is often a major cause for concern among
caregivers.
Understanding the relative influence of genetic and
environmental factors on FF at different ages is key
for informing intervention efforts to support
Zeynep Nas and Moritz Herle joint first authors.
Alison Fildes and Clare H. Llewellyn joint senior authors.
Conflict of interest statement: No conflicts declared.
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for Child and
Adolescent Mental Health.
This is an open access article under the terms of the Creative Commons Attribution License , which permits use, distribution and reproduction in any
medium, provided the original work is properly cited.
Journal of Child Psychology and Psychiatry 66:2 (2025), pp 241–252 doi:10.1111/jcpp.14053

caregivers in managing fussy eating behaviours. For
example, if environmental influences are stronger at
one-time point, then interventions targeting this age
group are likely to have the greatest chance of
success. On the other hand, if genetic influences
dominate at a particular age, then interventions may
be more challenging, targeted, personalised, and
more intensive management may be needed, while
reassuring parents that they are not to blame. We
have previously conducted cross-sectional twin
studies of FF in toddlerhood (16 months) and early
childhood (3 years) (Fildes, van Jaarsveld, Cooke,
Wardle, & Llewellyn, 2016; Smith et al., 2017), but to
date there are no studies of the relative influence of
genetic and environmental factors on the develop-
mental course of FF from toddlerhood through to
adolescence. In this context, little is known about the
aetiology of problematic fussy eating patterns char-
acterised by continued high levels of fussy eating
from toddlerhood into adolescence, captured in
trajectory analyses, or the changing genetic and
environmental influences on FF at different develop-
mental stages.
The main aims of this study are to:
1 Model the developmental trajectory of fussy eating
from toddlerhood to early adolescence and esti-
mate genetic and environmental contributions to
individual differences in trajectory parameters.
2 Estimate genetic and environmental contributions
to fussy eating at different ages from toddlerhood
to early adolescence and examine stability and
change in these relative contributions to fussy
eating from 16 months to 13 years of age.
Methods
Sample
Participants were from Gemini, a population-based cohort of
twin children born in England and Wales in 2007 (Jaarsveld,
Johnson, Llewellyn, & Wardle, 2010), who have been followed
up for over a decade. Gemini participants were recruited
through the UK Office for National Statistics, which contacted
all eligible families with twins born between March and
December 2007 (n=6,754), for consent to participate. Of the
3,435 families who consented, 2,402 completed the baseline
questionnaire, and these families constitute the main baseline
Gemini cohort. The cohort includes measures capturing
growth, eating behaviours, appetite, home environment, and
health outcomes. The study was initially granted ethical
approval in 2007 through the University College London
Committee for the Ethics of non-National Health Service
Human Research, with continuing approval for subsequent
data collection waves. In this study, we use data collected from
parents when the twins were on average 16 months
(n=3,854), 3 years (n=2,666), 5 years (n=2,098), 7 years
(n=703) and 13 years of age (n=970).
Measures
Fussy eating. ‘FF’ is a subscale of the parent-reported
Child Eating Behaviour Questionnaire (CEBQ) a widely used
parent-reported measure of fussy eating behaviours in chil-
dren. It has good internal and test–retest reliability (Wardle,
Guthrie, Sanderson, & Rapoport, 2001) and has been validated
against psychiatric interviews (Steinsbekk, Sveen, Fildes,
Llewellyn, & Wichstrøm, 2017), clinical measures of feeding
problems (Rogers, Ramsay, & Blissett, 2018) and independent
behavioural observation measures of fussy eating in children
(Rendall, Dodd, & Harvey, 2020). The scale incorporates both
FF (two items) as well as food neophobia (four items). The two
constructs are highly correlated (r=.72, p<.001) and share a
common aetiology as previously shown in Gemini at
16 months (Smith et al., 2017). The scale was administered
when the children were 16 months, 3, 5, 7, and 13 years old
(Wardle et al., 2001). At 7 years, data were only collected in a
subsample as part of a targeted round of dietary data
collection. The subscale consists of six items and is rated on
a five-point Likert scale from ‘never’ to ‘always’ (example item:
‘My child decides that s/he doesn’t like a food, even without
tasting it’). At 16 months, an adapted version of the CEBQ (the
CEBQ-T), suitable for toddlers was used, although the items
for the FF subscale remain identical. Items included are listed
in Table S1.
Zygosity, age and sex. Opposite-sex pairs were classi-
fied as dizygotic. The zygosity of same-sex twin pairs was
assigned using a widely used parent-reported questionnaire
measure of similarity (Herle, Fildes, van Jaarsveld, Rijsdijk, &
Llewellyn, 2016; Price et al., 2000) that was completed at
8 months and again at 29 months and validated using DNA
(for more detail on the process, see Herle et al., 2016). Child
sex was parent-reported at baseline, and age of the twins at
each wave was calculated from the parent-reported date of
birth and the date the questionnaires were completed at the
included waves.
Data analyses
Longitudinal change in fussy eating –developmen-
tal trajectory analysis. We first examined the overall
trajectory of fussy eating across the five time points using
linear mixed model analysis in Stata. The mixed-effects
framework lends itself to the analyses of repeated measures
as it accounts for the nonindependence of measures within an
individual. The mixed model estimated three parameters:
intercept, linear slope, and quadratic slope. Briefly, the
intercept refers to the starting average FF score for the sample,
the linear slope describes the pattern of change from the
intercept, and the quadratic slope indicates whether this linear
pattern changes over time. After model fitting, we extracted
parameters using the best linear unbiased predictions
(BLUPs). Genetic analyses quantifying the variance and
covariance between the three parameters were undertaken
using the twin design, as outlined in more detail below.
Analyses were pre-registered, see here: https://osf.io/
jcwsq. We diverged from the pre-registration in that we
originally aimed to derive multiple latent trajectories of fussy
eating using a growth mixture model approach. However,
models did not converge –potentially due to reduced variability
within twin pairs, and hence we were not able to derive
meaningful latent trajectories. Instead, we estimated one
average trajectory using a mixed model.
Genetic analyses. The twin design is based on the known
genetic difference between monozygotic (MZ; identical) and
dizygotic (DZ; nonidentical) twin pairs; MZ pairs share 100% of
their genetic material, whereas DZ pairs share, on average,
50% of their segregating DNA. Using this information, the total
phenotypic variance of a trait can be decomposed into three
latent components: Additive genetic influences (A); common/
shared environmental influences (C); and unique
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
242 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52

environmental influences (E), which also includes random
measurement error. The pattern of twin correlations provides
an indication of the genetic and environmental influences on
variation in, and covariation between, traits; maximum
likelihood structural equation modelling (MLSEM) is used to
derive more precise estimates of A, C, and E with 95%
confidence intervals and goodness-of-fit statistics.
Twin correlations
Comparisons between correlations for MZ and DZ pairs at each
time point provide an indication of the aetiology of the trait at
each age. If MZ twin correlations are larger than the DZ twin
correlations this indicates a genetic contribution, with larger
differences suggesting a larger contribution of genetic differ-
ences to the variance in that trait. On the other hand,
similar-sized correlations between MZ and DZ twins indicate
an important contribution from the shared environment and
less genetic influence. The extent to which the MZ correlation is
not 1.0 indicates the contribution of unique environmental
factors (and random measurement error), which are the only
source of differences between MZ twin pairs (because they are
matched completely for genetic and shared environmental
influences). Within-twin cross-time correlations indicate the
phenotypic stability in the measured phenotype (e.g. FF) over
time (r
pheno
). Cross-twin, cross-time correlations provide an
indication of the likely source of covariance across ages (i.e. the
extent to which the same genetic or environmental factors
contribute to longitudinal stability in FF). In the same way,
cross-twin, cross-trait correlations provide an indication of the
extent to which common genetic or environmental factors
contribute to covariation between different traits measured at
the same time point (i.e. different trajectory parameters of FF).
Maximum likelihood structural equation modelling
We fitted a multivariate Correlated Factors Model to examine
the genetic and environmental contributions to variation in
and covariation between the three parameters of FF (FF)
trajectories (intercept, slope, and quadratic terms). This model
is appropriate for examining shared aetiology underlying
multiple variables that are measured cross-sectionally, with
no particular temporal ordering between them. In this model,
the aetiological correlations (additive genetic correlation; rA,
common/shared environmental correlation; rC and unique
environmental correlation; rE) indicate the extent to which the
genetic and environmental influences underlying the different
phenotypes are the same. They can be interpreted like a
Pearson’s correlation: for example, a high positive correlation
indicates that many of the same genetic influences that
contribute to higher scores on one trait, also influence higher
scores on the other trait; a high negative genetic correlation
indicates that many of the same genetic influences that
contribute to higher scores on one trait, also influence lower
scores on the other trait.
We fitted a longitudinal Cholesky Decomposition Model to
examine genetic and environmental influences on variation in
FF at each of the five ages, and to stability and change in FF
over time. A Cholesky Decomposition Model is appropriate for
examining the variation in and covariation between longitudi-
nal data –i.e. the same phenotype that has been measured
repeatedly at different time points (Rijsdijk & Sham, 2002). A
longitudinal ACE Cholesky model is conceptually comparable
to a hierarchical regression in that the independent contribu-
tions of genetic and environmental influences are assessed
after the contributions of previous influences have been
accounted for. In the Cholesky decomposition, these are
known as unique genetic and environmental influences at
each age, independent of influences from previous time points
(these effects are estimated by paths a11, a22, a33, c11, c22,
..., e11, e22, as demonstrated on the example path diagram
shown in Figure S1). Additionally, the method estimates the
extent to which genetic and environmental influences carry
over from one time point to another (these are known as
overlapping aetiological effects, and are estimated by paths
a21, a31, a41, etc.). The first aetiological influences (i.e. paths
a11, c11, and e11) represent all genetic and environmental
influences on the first measured age point (these capture
influences that are both unique to this age, as well as those
overlapping with previous ages, but for simplicity, we represent
these as unique influences in this study).
Genetic analyses were conducted in R using the structural
equation modelling package, OpenMX (Neale et al., 2016). The
program handles missing data via full-information
maximum-likelihood estimation (FIML). Sex and ages corre-
sponding to each wave were regressed out from raw data prior
to twin modelling analyses, and residuals were analysed. We
began by fitting a fully saturated phenotypic model (Gaussian
decomposition), which estimated all parameters without
constraints. Next, we fitted two submodels, the first constrain-
ing means and variances across twin order and zygosity, and
the second also constraining the phenotypic correlations to be
equated across twin order and zygosity specifying symmetric
cross-twin cross-trait correlation matrices in MZ and DZ
groups. We also extracted the most parsimonious twin model
by dropping non-significant parameters. For instance, if we
find that shared environmental effects are non-significant, we
will constrain these non-significant parameters to zero and test
whether this results in a significant reduction in fit.
The goodness-of-fit of the models was assessed with minus
twice the log-likelihood (2LL). Difference in 2LL between a
full model and a nested submodel (simpler model with fewer
parameters) was assessed by v
2
tests and p-values, such that
the more parsimonious nested model is preferred only if this
does not result in a significant reduction in fit. In addition, we
used the Akaike Information Criterion (AIC) and Bayesian
Information Criterion as indicators for model fit, with lower
values indicating the better fitting model (Posada &
Buckley, 2004).
Results
Descriptive statistics on child (zygosity, age, sex and
FF score) and family characteristics (mothers’ age at
birth, ethnicity, education level at baseline, house-
hold income category, and twins’ gestational age) at
each time wave are provided in Table 1. Nonresponse
analyses between the baseline Gemini cohort
(8 months) suggest that the 7- and 13-year data
are of higher socio-economic status (SES), feature a
larger proportion of White families and mothers are
generally older and have a longer gestational period
(Table S11). The distribution of FF at the five time
points as well as a spaghetti plot can also be viewed
in the Supporting Information (Figures S2 and S3).
There was no significant reduction in fit from the
fully saturated model to the more parsimonious
Cholesky and Correlated Factors Models, as indi-
cated by non-significant changes in v
2
and decreas-
ing AIC values (Tables S2 and S3).
Genetic and environmental influences on food
fussiness (FF) trajectory parameters
Food fussiness trajectory. Figure 1depicts the
predicted single trajectory of FF across the five time
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.14053 Longitudinal examination of FF 243

points, derived from the mixed effects model.
Descriptive statistics for the trajectory parameters
(intercept, linear and quadratic slopes), along with
phenotypic and MZ-DZ correlations can be found in
Supporting Information (Tables S4–S6). The inter-
cept was positively correlated with the linear slope
(r=.67, 95% CI: 0.65, 0.69), indicating that chil-
dren who start at a higher average FF score tend to
have a larger increase in FF over time. There was a
negative correlation between the intercept and
quadratic slope, indicating that children who
started at a higher FF score also had a steeper
decrease of FF over time (r=.82, 95% CI: 0.83,
0.81). There was also a negative correlation
between the linear slope and quadratic slope,
indicating that children who had a higher linear
slope also showed a steeper decrease of FF over
time (r=.95, 95% CI =0.95, 0.94). These
findings suggest that those who started at a higher
FF score and those with a higher linear increase
also showed steeper decreases in fussiness from 7
to 13 years in this trajectory pattern, though
fussiness scores remained higher than at starting
point on average, highlighting the persistence of FF
throughout this period.
Twin correlations: Cross-twin, within trait correla-
tions for MZ pairs for each of the trajectory param-
eters, were more than twice the size of the DZ pair
correlations, suggesting a large genetic contribution
to all three parameters. This pattern was similar to
the cross-twin cross-trait correlations, indicating
that the associations between the three trajectory
parameters are likely to be genetically driven.
Correlated factors model: We found high heritabil-
ity for all three parameters (Table S7), indicating that
the variance in each aspect of the single trajectory of
FF is under strong genetic influence. Figure 2shows
the Correlated Factors Model with heritability esti-
mates and aetiological correlations between trajec-
tory parameters (Table S8). We found a large,
positive genetic correlation between intercept and
linear slope, indicating that many of the same genes
that influence a higher starting average FF score also
influenced the greater increases from toddlerhood to
Table 1 Descriptive statistics
Child variable
Wave
16 months 3 years 5 years 7 years 13 years
n3,854 2,666 2,098 703 970
Zygosity MZ =1,232
DZ =2,562
Unknown =60
MZ =909
DZ =1,731
Unknown =26
MZ =700
DZ =1,384
Unknown =14
MZ =232
DZ =471
MZ =334
DZ =630
Unknown =6
Mean child age (SD) 15.82 months
(1.15)
3.46 (0.27) 5.15 (0.13) 7.18
(0.24)
12.90 (0.71)
Sex (% male) 1,909 (49.5%) 1,321 (49.5%) 1,030 (49.1%) 349
(49.6%)
474 (48.9%)
FF average score (SD) (range 1–5) 2.17 (0.70) 2.66 (0.84) 2.77 (0.83) 2.67
(0.81)
2.60 (0.89)
Family variable
Average annual household income category
(1 =Low, 2 =Medium, 3 =High) (%)
1=1,114
(29.8)
2=1,794
(48.1)
3=826 (22.1)
1=701 (27.2)
2=1,269
(49.1)
3=612 (23.7)
1=546 (26.9)
2=998 (49.1)
3=488 (24.0)
1=145
(21.1)
2=354
(51.5)
3=188
(27.4)
1=208
(21.9)
2=484
(51.1)
3=256
(27.0)
Mother’s average age at birth in years (SD) 33.35 (5.05) 33.62 (4.75) 33.84 (4.76) 34.46
(4.51)
34.25 (4.36)
Gestational age average (weeks) (SD) 36.21 (2.47) 36.19 (2.51) 36.26 (2.43) 36.42
(2.32)
36.32 (2.52)
Mother’s education level (%)
Low 740 (19.2) 436 (16.4) 336 (16.0) 59 (8.4) 122 (12.6)
Intermediate 1,378 (35.8) 939 (35.2) 726 (34.6) 206
(29.3)
294 (30.3)
High 1,736 (45.0) 1,291 (48.4) 1,036 (49.4) 438
(62.3)
554 (57.1)
Mothers’ ethnicity (%)
White 3,636 (94.3) 2,534 (95.0) 2,004 (95.5) 669
(95.2)
932 (96.1)
Non-white 218 (5.7) 132 (5.0) 94 (4.5) 34 (4.8) 38 (3.9)
Income was measured at child aged ~8 months and categorised as follows (high =>£67.5 k high income; medium =£30 k–67.5 k
average UK income, low =<£30 k less than average UK income). Maternal education was measured at child age ~8 months and
categorised as: low =no qualifications or high school education for example CSE, GCSE, O level; intermediate =vocational
qualification or advanced high school education, and high =University-level education.
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
244 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52

middle childhood. The genetic correlations between
intercept and quadratic slope, and between linear
and quadratic slopes were strong and negative,
suggesting that many of the same genes that
influence higher starting average FF scores, and
higher linear increases from toddlerhood to middle
childhood also contributed to a steeper decrease in
FF from middle childhood to adolescence. In other
words, there is a large genetic influence on both the
stability and any trajectory changes observed in
fussy eating. Unique environmental correlations
were of similar magnitude and in the same pattern
of direction.
Genetic and environmental influences on variation
in and covariation between food fussiness at
each age
Twin correlations. Phenotypic correlations
between each time point were positive and signifi-
cant and ranged from moderate to large
(r
pheno
=.30–.77) (Table 2), indicating that FF is a
moderately to highly stable trait across development.
Cross-twin, within-time correlations, indicated that
genetic influences become more prominent over time
because the size of the difference between the MZ
and DZ correlations became larger as children
matured. Cross-twin, cross-time correlations sug-
gested that continuing genetic influences contribute
to stability in FF over time, as indicated by larger
correlations for MZ than DZ twins (Table S9).
Cholesky decomposition model. We fitted a longi-
tudinal ACE Cholesky Decomposition Model for FF
at the five time points. To extract the most parsimo-
nious model, we dropped C paths for ages 3, 5, 7,
and 13 years, as these were small and non-
significant. Constraining these paths to zero did
not result in a significant reduction in fit (Table S3).
However, we retained the C path for FF at
16 months, as this estimate was sizeable and
significant (ACE estimates for the full unconstrained
model can be viewed in Table S10).
The total contribution of genetic influences on
variation in FF (heritability) ranged from 60% to 84%
at the five measurement ages, with heritability at
16 months being significantly lower than at all
subsequent ages (Table 3). The shared environment
was only significant at 16 months and explained
25% of the individual differences in FF at this age.
Unique environmental influences explained 15%–
26% of the total individual differences in FF across
the five measurement points.
The Cholesky Decomposition Model also indicated
the extent to which the genetic and environmental
influences on variation in FF at 3, 5, 7, and
13 years were unique to each age or overlapped
with previous ages. Figure 3summarises the total
unique and overlapping A and E influences at each
age; Figures S4 and S5 show the path estimates
from the fully constrained Cholesky Decomposition
Model for A and E influences, respectively. The
proportion of genetic influence that was unique to
each age decreased over time, while overlapping
genetic influences increased cumulatively as chil-
dren grew older, indicating that common genetic
influences contribute strongly to stability in FF over
time, in line with the pattern of twin, cross-twin,
cross-time correlations. Of note, the genetic covari-
ation path between 16 months and 13 years (i.e.
the extent to which the genetic influences on FF in
adolescence are the same as those already
Figure 1 Longitudinal trajectory of mean food fussiness in the Gemini sample from 16 months to 13 years of age
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.14053 Longitudinal examination of FF 245

expressed in toddlerhood) was moderate (A
Chol =0.20, 95% CI =0.13, 0.29) indicating that
the genetic influences on FF are both stable and
persist over a period of greater than 10 years
(Figure S4). However, there was also substantial
unique genetic influence at 13 years of age, sug-
gesting new genetic influences on FF come online as
the twins reach adolescence. Non-shared environ-
mental influences also increased over time and
influences carried over from previous ages contrib-
uted to the stability of FF from 5 to 13 years
(overlapping influences were not observed between
toddlerhood and 3 years of age) (Figure S5). How-
ever, overall, common genetic factors played a far
more important role in contributing to stability in
FF over time than common unique environmental
factors. This was evident from the considerably
larger % of total variance in FF at each age
explained by overlapping genetic versus overlapping
unique environmental factors (3 years: 32% vs. 0%;
5 years: 48% vs. 4%; 7 years: 52% vs. 10%;
13 years: 43% vs. 14%).
Figure 2 Correlated Factors ACE Model of food fussiness trajectory parameters, including intercept, linear, and quadratic slopes. For
simplicity, shared environmental correlations were not reported in this figure, as they were very small and largely non-significant. Full
details of this model can be found in Supporting Information
Table 2 Pairwise correlation correlations of FF scores (95% CI) measured using the CEBQ
16 months 3 years 5 years 7 years
3 years 0.45 (0.42, 0.49)
5 years 0.41 (0.37, 0.45) 0.71 (0.68, 0.73)
7 years 0.35 (0.29, 0.40) 0.66 (0.62, 0.69) 0.77 (0.74, 0.80)
13 years 0.30 (0.24, 0.36) 0.52 (0.46, 0.56) 0.63 (0.58, 0.67) 0.70 (0.65, 0.74)
16 months n=3,854; 3 years n=2,666, 5 years n=2,098, 7 years n=703; 13 years n=970.
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
246 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52

Discussion
This is the first study to establish the genetic and
environmental contributions to an overall FF trajec-
tory spanning from toddlerhood to early adolescence
and to quantify genetic and environmental influ-
ences on individual differences in trait stability and
change across five time points.
Summary of main findings
Modelling of a single longitudinal trajectory revealed
FF to be, on average, a highly stable trait. The three
trajectory parameters (intercept: i.e. starting average
score; linear slope: i.e. rate of linear change; and
quadratic slope: i.e. change in the linear growth)
were strongly correlated and all under strong genetic
influence, with 74%–79% of individual differences
explained by genetic factors. Most of the remaining
variance in these trajectory parameters was due to
non-shared environmental influences (and measure-
ment error: 13%–25%), with small and mostly non-
significant contributions from the shared environ-
ment. We found significant and large genetic and
unique environmental correlations between trajec-
tory components of FF. Specifically, there were large
and positive genetic and unique environmental
correlations between the intercept and linear slope,
indicating that the same genetic and unique envi-
ronmental influences that contributed to a higher
average FF score in toddlerhood also contributed to a
greater linear increase in FF from toddlerhood to
middle childhood. On the other hand, there were
large negative aetiological correlations between the
linear and quadratic slope, and between the inter-
cept and quadratic slope. This indicated that many
of the same genetic and unique environmental
influences that contributed to a higher average FF
score and greater linear increases in FF from
toddlerhood to middle childhood also contributed
to the decline in FF from middle childhood to early
adolescence. In the context of very high estimates for
all three trajectory parameters, these results point
towards shared genetic effects being largely respon-
sible for the continuity and persistence of FF
throughout childhood and into adolescence, includ-
ing patterns of persistent and enduring FF.
Discrete age analyses indicated moderate–large
phenotypic associations across time, indicating, that
FF is a moderately to highly stable trait during child
development, in line with the trajectory analyses. We
also showed a high heritability of FF at every age
from toddlerhood to early adolescence, with the
lowest genetic contribution in toddlerhood. At
16 months, genetic factors explained 60% of the
variance in FF, which had risen to 74%–84%
between the ages of three and 13. Persisting genetic
influences that endured throughout development
were largely responsible for stability in FF over time.
Genetic effects already in operation in toddlerhood
continued to have an impact over 10 years later,
evidenced by overlapping genetic influences from
16 months to 13 years. In fact, from 5 years of age
onwards, about half of the total variance in FF was
explained by genetic effects ‘carried over’ from
previous time points, with the biggest contribution
observed at 7 years. We found that unique environ-
mental influences increased over time, as children
transitioned towards greater independence. On the
other hand, shared environmental influences were
only significant in toddlerhood (25% of variance
explained) and were largely negligible by age three.
Findings in context
Our findings are in line with previous work suggest-
ing that fussy eating and related behaviours can be
persistent and enduring for some children (Cardona
Cano et al., 2015; Mascola et al., 2010) in line with
theory and previous epidemiological studies (Taylor
et al., 2015). However, the average trajectory scores
never dipped below the initial starting average score,
indicating children continued to display more FF at
13 years old than they had at 16 months. Previous
work also suggested fussy eating increases and then
decreases between 7 and 13 years, though fussy
eating scores reached baseline levels at early ado-
lescence, suggesting the need to explore this devel-
opmental period further (Carter Leno et al., 2022).
The results from the longitudinal twin modelling
align with previous cross-sectional work in Gemini
reporting moderate heritability estimate for FF of
46% at 16 months of age (Smith et al., 2017), and a
high heritability estimate of 78% by 3 years of age
(Fildes et al., 2016). The current study expands on
this earlier work, by embedding a longitudinal
approach within a behaviour genetics design. Previ-
ous longitudinal work has mainly been phenotypic,
such as the association of FF as a risk factor for
behaviours like disordered eating, diet quality, and
weight status (Antoniou et al., 2016; da Costa
et al., 2022; Herle, Stavola, et al., 2020). This study
is the first to focus specifically on fussy eating, using
Table 3 Total variance explained by genetic and environmen-
tal influences across time
Time point A C E
16 months 0.60 (0.53, 0.67) 0.25 (0.18,
0.31)
0.15 (0.14, 0.17)
3 years 0.83 (0.81, 0.86) –0.17 (0.14, 0.19)
5 years 0.84 (0.80, 0.86) –0.16 (0.14, 0.20)
7 years 0.77 (0.70, 0.82) –0.23 (0.18, 0.30)
13 years 0.74 (0.67, 0.80) –0.26 (0.20, 0.33)
Note:A=Additive genetic influences; C =Common environ-
mental influences; E =Unique environmental influences
(including measurement error). Results are from constrained
ACE model. C influences were dropped due to small, non-
significant estimates at 3,5,7, and 12 years. The total variance
in a trait includes both unique influences at that given age as
well as influences carried over from previous age(s).
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.14053 Longitudinal examination of FF 247

repeated measures spanning more than a decade,
and offers a unique behaviour genetic context.
The shared environment being significant only in
toddlerhood is expected, in line with previous Gemini
research (Smith et al., 2017). Another British cohort
study indicated that shared environmental influ-
ences on variation in food preferences become
displaced, and mostly accounted for, by genetics
and the unique environment in adulthood (Smith
et al., 2016). Research suggests high heritability of
ARFID (79%), which features selective eating as a
common component, with no significant effect of the
shared environment (Dinkler et al., 2023). The high
genetic component of FF aligns with that observed
for other appetitive traits such as food responsive-
ness and satiety responsiveness in infancy and late
childhood (Carnell, Haworth, Plomin, & War-
dle, 2008; Llewellyn, van Jaarsveld, Johnson,
Carnell, & Wardle, 2010). For broader context, this
finding is also in the range reported by studies of
disordered eating in adolescents suggesting
moderate–high heritability estimates varying
between 28% and 83% (Thornton, Mazzeo, &
Bulik, 2011). Genetic differences in the population,
therefore, appear to play a particularly important
role in determining individual differences in many
different eating behaviour phenotypes in infancy,
toddlerhood, and adolescence.
Implications
These findings are important for our understanding
of fussy eating in children for three main reasons.
First, children’s fussy eating is a major cause of
concern for caregivers, who often blame themselves,
or are blamed by others, for their child’s restricted
32
48 52 43
60
51
36 25
31
0
10
20
30
40
50
60
70
80
90
FF16 FF3 FF5 FF7 FF13
Variance explained (%)
Food fussiness measurement age
Overlapping and unique genetic effects on food fussiness
(FF)
Unique genetic influence
at time point
Overlapping genetic
influence with previous
time point
4
10
14
15 17
12
13
12
0
5
10
15
20
25
30
FF16 FF3 FF5 FF7 FF13
Variance explained (%)
Food fussiness measurement age
Unique and overlapping non-shared environmental effects on
food fussiness (FF)
New unique environmental
influence at time point
Overlapping unique
environmental influence
with previous time point
Figure 3 X-axis represents the five measurements of food fussiness: FF16 =16 months, FF3 =3 years, FF5 =5 years, FF7 =7 years,
FF13 =13 years. Y-axis represents the proportion of variance explained by genetic and environmental influences. Shared environmental
influences were non-significant beyond toddlerhood and have, therefore, been omitted from this figure
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
248 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52

diets or food rejection. This study indicates that FF
is under strong genetic influence, which can remain
influential throughout childhood. This finding may
help to alleviate parental blame and explain why
siblings raised in the same environment often
express very different selective eating behaviours.
Second, the shared environment was only found to
significantly impact individual differences in fussy
eating behaviour in toddlerhood, indicating envi-
ronmental or family-based interventions targeting
children’s fussy eating behaviours (such as
repeated exposure and increasing the variety of
fruits and vegetables offered in the home), may be
most effective in the very early years (Kamarudin
et al., 2023). These findings do not imply that FF
cannot be changed in response to behavioural
interventions; however, they suggest that it may
be a more challenging behaviour to modify in
comparison to behaviours that are under predom-
inantly environmental influence. It is also worth
noting that the unique environment is found to be
stable and steadily increasing over time, implicating
that interventions for FF could be implemented
across childhood and into adolescence. Third, FF is
a stable trait as indicated by moderate-to-high
phenotypic correlations over time, and sizeable
continuing genetic influences driving stability
across development. Fussy eating behaviours are
not necessarily just a ‘phase’, but potentially follow
a persistent trajectory. These findings suggest that
toddlers who present with higher FF are also more
likely to experience greater increases in FF as they
mature. This is of interest not only to researchers
but also to clinicians and the wider child and
adolescent health community. Given the links
between fussy eating and other physical and
psychological health outcomes, including eating
disorders such as ARFID (Dovey, 2018), early
detection and intervention for FF in toddlerhood
may reduce the expression of this behaviour across
development.
Future directions
This is the most comprehensive longitudinal twin
study of fussy eating in childhood to date. We tested
a single trajectory for FF in this paper, and future
work may provide more refined insight into the
potential trajectory classes over time. There are
ongoing efforts to understand the genetic architec-
ture of fussy eating, including genome-wide analyses
(Abdulkadir et al., 2022). However, such studies
require large sample sizes and fussy eating –and
eating behaviours in general –are not commonly
measured in big cohort studies with genomic data,
making these approaches challenging. Another
important line of enquiry is the phenotypic and
genetic associations between fussy eating and other
neurodevelopmental disorders in childhood. Previ-
ous research has indicated that food neophobia (fear
of new foods, closely related to fussy eating) was
elevated in children with higher autistic-like traits
(Wallace, Llewellyn, Fildes, & Ronald, 2018) and that
fussy eating might be a mediator on the commonly
observed path between autism spectrum disorder
traits in childhood and risk for disordered eating in
adolescence (Carter Leno et al., 2022; Solmi
et al., 2021). The relationship between fussy eating
and other neurodevelopmental conditions, such as
ADHD and mood and anxiety disorders, has also
been documented and is worth exploring further
(Thorsteinsdottir, Olafsdottir, Brynjolfsdottir, Bjar-
nason, & Njardvik, 2022; Wu et al., 2023). Last,
these findings provide new insights into the com-
plexity of FF and the need for compassionate,
evidence-based interventions and guidance to sup-
port caregivers feeding children with varying levels
of FF.
Strengths and limitations
As with most longitudinal studies, attrition over time
can contribute to lower statistical power. In this
study, there were fewer participants at age seven
than at other ages as only a subsample of highly
engaged participants were invited to take part in this
wave of data collection; this could have resulted in a
change in the variance–covariance structure in the
multivariate Cholesky model. Although the study
spans a large age range, FF can also shift throughout
adolescence and into adulthood, prompting replica-
tion at other age points. Although the FF items
remained identical throughout the different time
points, we acknowledge that the scale measures
fussy eating at different developmental points and,
therefore, may capture a slightly different underlying
construct at each stage. We recommend that mea-
surement invariance in the scale is examined further
in future studies. Twin studies also rely on key
modelling assumptions, such as the ‘equal environ-
ments assumption’, which assumes that MZ and DZ
twins share environments to a similar extent, and
although debated, the assumption has generally
been supported (Felson, 2014). FF was
parent-reported and parents’ knowledge of their
twins’ zygosity might influence the ratings of their
twins’ behaviours, for example parents may be
inclined to rate their twins more similarly because
they believe them to be identical. However, we
examined this in Gemini using the misclassified
twin design and parental zygosity perception did not
influence their reporting of fussy eating (Herle
et al., 2016). The Gemini sample is fairly homoge-
nous, including a large proportion of White-British
households of higher socio-economic positions com-
pared to the general population. Most characteristics
at each wave remain similar descriptively, though,
upon further inspection, data at ages 7 and 13 years
are less representative of the baseline Gemini cohort.
We, therefore, suggest taking findings with this into
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.14053 Longitudinal examination of FF 249

consideration and recommend replication in more
diverse populations. Considering this, to meaning-
fully advance the field, future research requires
large-scale, genetically sensitive, and longitudinal
analyses of fussy eating. Replication is particularly
warranted in non-western populations where food
culture, parental feeding practices and food security
may vary considerably.
Conclusions
This novel longitudinal examination provides evi-
dence of FF being a highly heritable trait that is
relatively stable from toddlerhood into early adoles-
cence, with genetic influences largely responsible for
its continuity.
Supporting information
Additional supporting information may be found online
in the Supporting Information section at the end of the
article:
Figure S1. Path diagram for full Multivariate Cholesky
twin model.
Figure S2. Distributions of food fussiness.
Figure S3. Spaghetti plot of food fussiness across
waves.
Figure S4. Path diagram depicting genetic paths from
constrained longitudinal Cholesky model.
Figure S5. Path diagram depicting unique environmen-
tal paths from constrained longitudinal Cholesky
model.
Table S1. Food fussiness items.
Table S2. Correlated factors twin model fit statistics
(intercept, linear slope and quadratic slope).
Table S3. Longitudinal Cholesky model fit statistics
(discrete age analysis).
Table S4. Descriptive statistics for intercept, linear
slope and quadratic slope for food fussiness.
Table S5. MZ - DZ correlations obtained from corre-
lated factors twin model for intercept, linear slope, and
quadratic slope.
Table S6. Phenotypic correlations between intercept
and linear-quadratic slopes.
Table S7. ACE components for intercept, linear and
quadratic slopes.
Table S8. Aetiological correlations between intercept
and linear-quadratic slopes.
Table S9. Longitudinal Cholesky model MZ and DZ
correlations obtained from constrained phenotypic
model (Sub model 1b).
Table S10. Standardised ACE estimates at each wave
obtained from full longitudinal Cholesky model.
Table S11. Nonresponse analyses between baseline
Gemini cohort and 7- and 13-year data.
Acknowledgements
The authors are grateful to the participants of the
Gemini study for their continued participation in this
research. The authors have declared that they have no
competing or potential conflicts of interest.
Data availability statement
The data that support the findings of this study are
available from the corresponding author upon reason-
able request.
Correspondence
Clare H. Llewellyn, Department of Behavioural Science
& Health, University College London, 1-19 Torrington
Place, London WC1E 7HB, UK; Email: c.llewellyn@ucl.
ac.uk
Key points
•What’s known: FF is a heritable eating behaviour trait that is common in early childhood and can limit
dietary diversity. FF can be challenging for parents to manage and has important implications for health
and development, as well as psychosocial implications such as family well-being and psychological
distress.
•What’s new: FF is a stable trait that peaks in middle childhood and declines slightly thereafter. It endures
into early adolescence, with individual variation in FF remaining under moderate–large genetic
influence across development, from toddlerhood into adolescence.
•What’s relevant: Parents are not to blame for their children’s innate fussy eating behaviours.
Interventions targeting FF could start as early as toddlerhood and may need to be tailored and
intensive at different developmental time points.
References
Abdulkadir, M., Herle, M., Mulder, R., Harris, H., Jansen, P., &
Micali, N. (2022). Genome-wide association study of child-
hood fussy eating. European Neuropsychopharmacology,63,
e125–e126.
Antoniou, E.E., Roefs, A., Kremers, S.P.J., Jansen, A.,
Gubbels, J.S., Sleddens, E.F.C., & Thijs, C. (2016). Picky
eating and child weight status development: A longitudinal
study. Journal of Human Nutrition and Dietetics: The
Official Journal of the British Dietetic Association,29,
298–307.
Cardona Cano, S., Tiemeier, H., Van Hoeken, D., Tharner, A.,
Jaddoe, V.W.V., Hofman, A., ... & Hoek, H.W. (2015).
Trajectories of picky eating during childhood: A general
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
250 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52

population study. The International Journal of Eating
Disorders,48, 570–579.
Carnell, S., Haworth, C.M.A., Plomin, R., & Wardle, J. (2008).
Genetic influence on appetite in children. International
Journal of Obesity,32, 10.
Carter Leno, V., Micali, N., Bryant-Waugh, R., & Herle, M.
(2022). Associations between childhood autistic traits and
adolescent eating disorder behaviours are partially mediated
by fussy eating. European Eating Disorders Review: The
Journal of the Eating Disorders Association,30, 604–615.
da Costa, M.P., Severo, M., Oliveira, A., Lopes, C., Hether-
ington, M., & Vilela, S. (2022). Longitudinal bidirectional
relationship between children’s appetite and diet quality: A
prospective cohort study. Appetite,169, 105801.
de Barse, L.M., Tiemeier, H., Leermakers, E.T.M., Voortman,
T., Jaddoe, V.W.V., Edelson, L.R., ... & Jansen, P.W. (2015).
Longitudinal association between preschool fussy eating
and body composition at 6 years of age: The generation R
study. International Journal of Behavioral Nutrition and
Physical Activity,12, 153.
Dinkler, L., Wronski, M.-L., Lichtenstein, P., Lundstr€
om, S.,
Larsson, H., Micali, N., ... & Bulik, C.M. (2023). Etiology of
the broad avoidant restrictive food intake disorder pheno-
type in Swedish twins aged 6 to 12 years. JAMA Psychiatry,
80, 260–269.
Dovey, T.M. (2018). Avoidant/restrictive food intake disorder:
An eating disorder on a spectrum with food neophobia. In S.
Reilly (Ed.), Food neophobia (pp. 329–349). Cambridge, UK:
Woodhead Publishing.
Felson, J. (2014). What can we learn from twin studies? A
comprehensive evaluation of the equal environments
assumption. Social Science Research,43, 184–199.
Fildes, A., van Jaarsveld, C.H., Cooke, L., Wardle, J., &
Llewellyn, C.H. (2016). Common genetic architecture under-
lying young children’s FF and liking for vegetables and fruit.
The American Journal of Clinical Nutrition,103, 1099–1104.
Herle, M., De Stavola, B., H€
ubel, C., Ferreira, D.L.S., Abdulk-
adir, M., Yilmaz, Z., ... & Micali, N. (2020). Eating behavior
trajectories in the first ten years of life and their relationship
with BMI. International Journal of Obesity,44, 1766–1775.
Herle, M., Fildes, A., van Jaarsveld, C., Rijsdijk, F., &
Llewellyn, C.H. (2016). Parental reports of infant and child
eating behaviors are not affected by their beliefs about their
Twins’ Zygosity. Behavior Genetics,46, 763–771.
Herle, M., Stavola, B.D., H€
ubel, C., Abdulkadir, M., Ferreira,
D.S., Loos, R.J.F., ... & Micali, N. (2020). A longitudinal
study of eating behaviours in childhood and later eating
disorder behaviours and diagnoses. The British Journal of
Psychiatry: the Journal of Mental Science,216, 113–119.
Jaarsveld, C.H.M., Johnson, L., Llewellyn, C., & Wardle, J.
(2010). Gemini: A UK twin birth cohort with a focus on early
childhood weight trajectories, appetite and the family
environment. Twin Research and Human Genetics,13,72–
78.
Kamarudin, M.S., Shahril, M.R., Haron, H., Kadar, M., Safii,
N.S., & Hamzaid, N.H. (2023). Interventions for picky eaters
among typically developed children—A scoping review.
Nutrients,15, 242.
Lafraire, J., Rioux, C., Giboreau, A., & Picard, D. (2016). Food
rejections in children: Cognitive and social/environmental
factors involved in food neophobia and picky/fussy eating
behavior. Appetite,96, 347–357.
Llewellyn, C.H., van Jaarsveld, C.H.M., Johnson, L., Carnell,
S., & Wardle, J. (2010). Nature and nurture in infant
appetite: Analysis of the Gemini twin birth cohort. The
American Journal of Clinical Nutrition,91, 1172–1179.
Machado, B.C., Dias, P., Lima, V.S., Campos, J., & Gonc
ßalves,
S. (2016). Prevalence and correlates of picky eating in
preschool-aged children: A population-based study. Eating
Behaviors,22,16–21.
Mascola, A.J., Bryson, S.W., & Agras, W.S. (2010). Picky eating
during childhood: A longitudinal study to age 11 years.
Eating Behaviors,11, 253–257.
McClelland, J., Robinson, L., Potterton, R., Mountford, V., &
Schmidt, U. (2020). Symptom trajectories into eating
disorders: A systematic review of longitudinal, nonclinical
studies in children/adolescents. European Psychiatry,63,
e60.
Neale, M.C., Hunter, M.D., Pritikin, J.N., Zahery, M., Brick,
T.R., Kirkpatrick, R.M., ... & Boker, S.M. (2016). OpenMx
2.0: Extended structural equation and statistical modeling.
Psychometrika,81, 535–549.
Posada, D., & Buckley, T.R. (2004). Model selection and model
averaging in phylogenetics: Advantages of Akaike informa-
tion criterion and Bayesian approaches over likelihood ratio
tests. Systematic Biology,53, 793–808.
Price, T.S., Freeman, B., Craig, I., Petrill, S.A., Ebersole, L., &
Plomin, R. (2000). Infant zygosity can be assigned by
parental report questionnaire data. Twin Research: The
Official Journal of the International Society for Twin Studies,
3, 129–133.
Rendall, S., Dodd, H., & Harvey, K. (2020). Behavioural
validation of a parent-report measure of child FF. Appetite,
154, 104796.
Rijsdijk, F.V., & Sham, P.C. (2002). Analytic approaches to
twin data using structural equation models. Briefings in
Bioinformatics,3, 119–133.
Rogers, S., Ramsay, M., & Blissett, J. (2018). The Montreal
Children’s hospital feeding scale: Relationships with paren-
tal report of child eating behaviours and observed feeding
interactions. Appetite,125, 201–209.
Smith, A.D., Fildes, A., Cooke, L., Herle, M., Shakeshaft, N.,
Plomin, R., & Llewellyn, C. (2016). Genetic and environmen-
tal influences on food preferences in adolescence. The
American Journal of Clinical Nutrition,104, 446–453.
Smith, A.D., Herle, M., Fildes, A., Cooke, L., Steinsbekk, S., &
Llewellyn, C.H. (2017). Food fussiness and food neophobia
share a common etiology in early childhood. Journal of Child
Psychology and Psychiatry,58, 189–196.
Solmi, F., Bentivegna, F., Bould, H., Mandy, W., Kothari, R.,
Rai, D., ... & Lewis, G. (2021). Trajectories of autistic social
traits in childhood and adolescence and disordered eating
behaviours at age 14 years: A UK general population cohort
study. Journal of Child Psychology and Psychiatry,62,75–
85.
Steinsbekk, S., Sveen, T.H., Fildes, A., Llewellyn, C., &
Wichstrøm, L. (2017). Screening for pickiness –A validation
study. International Journal of Behavioral Nutrition and
Physical Activity,14,2.
Taylor, C.M., & Emmett, P.M. (2019). Picky eating in children:
Causes and consequences. The Proceedings of the Nutrition
Society,78, 161–169.
Taylor, C.M., Steer, C.D., Hays, N.P., & Emmett, P.M. (2019).
Growth and body composition in children who are picky
eaters: A longitudinal view. European Journal of Clinical
Nutrition,73, 869–878.
Taylor, C.M., Wernimont, S.M., Northstone, K., & Emmett,
P.M. (2015). Picky/fussy eating in children: Review of
definitions, assessment, prevalence and dietary intakes.
Appetite,95, 349–359.
Thornton, L.M., Mazzeo, S.E., & Bulik, C.M. (2011). The
heritability of eating disorders: Methods and current find-
ings. Current Topics in Behavioral Neurosciences,6, 141–
156.
Thorsteinsdottir, S., Olafsdottir, A.S., Brynjolfsdottir, B.,
Bjarnason, R., & Njardvik, U. (2022). Odds of fussy eating
are greater among children with obesity and anxiety. Obesity
Science and Practice,8,91–100.
Wallace, G.L., Llewellyn, C., Fildes, A., & Ronald, A. (2018).
Autism spectrum disorder and food neophobia: Clinical and
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
doi:10.1111/jcpp.14053 Longitudinal examination of FF 251

subclinical links. The American Journal of Clinical Nutrition,
108, 701–707.
Wardle, J., Guthrie, C.A., Sanderson, S., & Rapoport, L.
(2001). Development of the children’s eating behaviour
questionnaire. Journal of Child Psychology and Psychiatry,
42, 963–970.
Wolstenholme, H., Heary, C., & Kelly, C. (2019). Fussy eating
behaviours: Response patterns in families of school-aged
children. Appetite,136,93–102.
Wu, D., Chen, S., Lin, S., Huang, R., Li, R., Huang, Y., ... & Li,
X. (2023). Eating behavior-mediated association between
attention deficit hyperactivity disorder and body fat mass.
Heliyon,9, e13641.
Zimmerman, J., & Fisher, M. (2017). Avoidant/restrictive food
intake disorder (ARFID). Current Problems in Pediatric and
Adolescent Health Care,47,95–103.
Zucker, N., Copeland, W., Franz, L., Carpenter, K., Keeling, L.,
Angold, A., & Egger, H. (2015). Psychological and psycho-
social impairment in preschoolers with selective eating.
Pediatrics,136, e582–e590.
Accepted for publication: 14 June 2024
Ó2024 The Author(s). Journal of Child Psychology and Psychiatry published by John Wiley & Sons Ltd on behalf of Association for
Child and Adolescent Mental Health.
252 Zeynep Nas et al. J Child Psychol Psychiatr 2025; 66(2): 241–52