You are what you eat: diet shapes body composition, personality and behavioural stability
Han and Dingemanse BMC Evolutionary Biology
You are what you eat: diet shapes body composition, personality and behavioural stability
Chang S. Han 0 1
Niels J. Dingemanse 1
0 Current address: School of Biological Sciences, University of Queensland , St Lucia 4072 , Australia
1 Behavioural Ecology, Department of Biology, Ludwig-Maximilians University of Munich , Großhaderner Str. 2, 82152 Planegg-Martinsried , Germany
Background: Behavioural phenotypes vary within and among individuals. While early-life experiences have repeatedly been proposed to underpin interactions between these two hierarchical levels, the environmental factors causing such effects remain under-studied. We tested whether an individual's diet affected both its body composition, average behaviour (thereby causing among-individual variation or 'personality') and within-individual variability in behaviour and body weight (thereby causing among-individual differences in residual within-individual variance or 'stability'), using the Southern field cricket Gryllus bimaculatus as a model. We further asked whether effects of diet on the expression of these variance components were sex-specific. Methods: Manipulating both juvenile and adult diet in a full factorial design, individuals were put, in each life-stage, on a diet that was either relatively high in carbohydrates or relatively high in protein. We subsequently measured the expression of multiple behavioural (exploration, aggression and mating activity) and morphological traits (body weight and lipid mass) during adulthood. Results: Dietary history affected both average phenotype and level of within-individual variability: males raised as juveniles on high-protein diets were heavier, more aggressive, more active during mating, and behaviourally less stable, than conspecifics raised on high-carbohydrate diets. Females preferred more protein in their diet compared to males, and dietary history affected average phenotype and within-individual variability in a sex-specific manner: individuals raised on high-protein diets were behaviourally less stable in their aggressiveness but this effect was only present in males. Diet also influenced individual differences in male body weight, but within-individual variance in female body weight. Discussion: This study thereby provides experimental evidence that dietary history explains both heterogeneous residual within-individual variance (i.e., individual variation in 'behavioural stability') and individual differences in average behaviour (i.e., 'personality'), though dietary effects were notably trait-specific. These findings call for future studies integrating proximate and ultimate perspectives on the role of diet in the evolution of repeatedly expressed traits, such as behaviour and body weight.
Behavioural stability; Developmental plasticity; Diet; Personality; Repeatability; Heterogeneous residual withinindividual variance
Behavioural ecologists increasingly focus on why
individuals differ in average behaviour (i.e., why there is
among-individual variation or ‘animal personality’) and
behavioural plasticity (i.e., why within-individual variation
or ‘behavioural stability’ differs among individuals) [1–3].
Quantitative genetics studies imply that, on average, 50%
of this individual variation in behaviour is due to additive
genetic effects (reviewed in ). Environmental factors
that permanently affect the phenotype therefore likely play
an equally important role in shaping individual behaviour.
Indeed, various studies have experimentally demonstrated
the importance of the early-life environment in
permanently shaping behavioural phenotypes [5–9]. For example,
studies on birds imply that food availability during
early-life can shape both aggressiveness and exploratory
tendency in adulthood .
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Empirical studies increasingly quantify the contribution
of among- and within-individual variation in shaping
phenotypic variation observed in animal behaviour [3, 11].
Recent research indicates that the (relative) magnitudes of
these two variance components can show sex-specificity
and spatiotemporal variation within a single species
[12–17]. This is in line with quantitative genetics studies
showing that the expression of genetic variance is often a
function of the environment . For example, the
additive genetic variance may either increase, or decrease,
under more favourable (or less ‘stressful’) conditions
[19, 20]. Behavioural examples are relatively few but
include work on the house cricket (Acheta domesticus)
showing that low quality diet increases within-individual
stability in anti-predatory behaviour, and consequently
increases its repeatability . Similarly, adults of the
western trilling crickets (Gryllus integer) exposed to bacterial
pathogens during ontogeny are less repeatable (due to
decreased among-individual variance) in boldness compared
to controls . Examples of environment-specific
withinindividual variation, furthermore, include work on
redwinged blackbirds (Agelaius phoeniceus), where females
decrease their within-individual variance in nestling
provisioning effort with increasing nestling age , studies on
hermit crabs (Pagurus bernhardus), where individuals
increase their level of within-individual behavioural
variability when faced with increased perceived predation risk
, and studies on zebra finches (Taeniopygia guttata),
where early dietary restriction decreases the amount of
within-individual variance in general activity .
Here we propose that the nutritional environment
represents a key environmental factor determining the
development and expression of behaviour, thus shaping
personality and level of behavioural stability . We
further propose that there are three components of
behaviour that can be affected by nutritional history. First,
average level of behaviour may vary among individuals
with different nutritional histories (represented by the
dotted line in Fig. 1). Indeed, previous studies have
demonstrated that poor nutrition can impair behavioural
development and provide reliable cues to the expression
of later behaviour [14, 22–28].
Second, the extent of individual differentiation (i.e.,
the amount of among-individual variance; VI) may differ
among groups of individuals differing in nutritional history
(variance component VI in Fig. 1). For example, under
high-quality dietary conditions (characterized by abundant
resources or balanced nutrients), multiple strategies with
equal fitness payoffs may exist by which resources are
allocated among costly behaviours, while this might not
be the case under low-quality conditions. If so, one
would expect the among-individual variance in behaviour
to be increased under high-quality dietary conditions 
(Fig. 1). One may, by contrast, also expect the opposite
Fig. 1 Schematic representation of how mean level, and within- and
among individual variances were expected to differ between
environments. Each dot with vertical line represents one theoretical
individual. The black line represents the extent of variation in
behaviour that is observed across observations of a single individual
(i.e., within-individual variance, VR). The change in the length of black
lines indicates the change in within-individual variance (VR) across
environments. The variation among dots represents the variation among
individuals in average phenotype (i.e., among-individual variance, VI). The
dashed line indicates a change in population-level mean
pattern, for example, when low-quality dietary
environments represent relatively novel environmental conditions
where additional (‘cryptic’) genetic variance is expressed
, leading to decreased among-individual variance under
high-quality dietary conditions (e.g., [14, 30, 31]).
Third, behavioural instability (i.e., the amount of residual
within-individual variance; VR) may also differ between
individuals subjected to different nutritional histories
(variance component VR in Fig. 1). Owing to the costs
associated with phenotypic plasticity , for example,
only individuals experiencing high-quality nutritional
environments might be able to invest in the sensory and
processing apparatus necessary for expressing adaptive
phenotypic plasticity. High-quality nutritional
environments during development might therefore enable
individuals to more flexibly respond to micro-environmental
variation experienced during adulthood. For example,
individuals subjected to high-quality nutritional
environments during development might develop the capabilities
to adjust their level of aggressiveness in response to changes
in their social environment, whereas individuals with a
lowquality dietary past might remain less flexible [33, 34]. This
multitude of possible effects on behavioural variance
components may thereby result in behavioural repeatabilities
that vary between treatment groups [14, 28]. Importantly,
as variances often increase as a function of the mean ,
effects of diet on mean behavioural phenotype are likely
coupled with those on variance components (as illustrated
in Fig. 1).
We further predict that effects of diet should differ
between sexes and life stages. Sex-differences in sensitivity
towards nutrient deficiency are expected because males
and females often differ in their nutritional preference
[26, 36–39]. In addition, as nutritional requirements
typically change throughout an individual’s life time ,
diets experienced during early vs. late life are predicted
to additively or interactively affect the expression of
phenotypes in adulthood (e.g. [41–44]). In the butterfly
Harmonia axyridis, for example, females raised on a
poor diet during adulthood showed decreased
fecundity only when not having experienced a poor diet as a
Here, we assessed how nutritional environments
experienced during juvenile and adult life affected various
behaviours and their variance components in males and
females of the Southern field cricket (G. bimaculatus).
Field crickets are well suited for testing the role of the
nutritional environment in shaping personality and
plasticity. Previous research on crickets has, for example,
shown that the carbohydrate:protein (C:P) ratio of a diet
affects the expression of morphology and reproductive
behaviours [22, 25, 26, 44–46]. This is not surprising as
protein is required for somatic development of nymphs,
and eggs produced by females. Similarly, carbohydrate is
needed to fuel general activity and male courtship
behaviour. In our experiments, we therefore manipulated
diets. We used a two-way factorial design with two
juvenile diets (high-carbohydrate versus high-protein) and
two adult diets (high-carbohydrate versus high-protein),
and measured the expression of multiple behavioural
(exploration, aggression and mating activity) and
morphological (body weight and lipid mass) traits. Given the
documented strong effects of the carbohydrate:protein
(C:P) ratios on various key phenotypic traits [22, 25, 26,
44–46], dietary environments are generally predicted to
alter body weight and the expression of behaviour,
though we appreciate that such effects may also interact
with social and non-social environmental factors, such
as the amount of competition for resources or mates.
Moreover, effects of diet on the expression of variance
components (i.e., among- or within-individual variances)
may also affect behavioural repeatability. We thus
measured 1) nutritional preferences of juveniles, adult males
and adult females and 2) nutritional intakes of juveniles
and adults faced with imbalanced diets (namely,
highcarbohydrate diets vs. high-protein diets, detailed below)
(experiment 1). We then assessed whether 3)
populationlevel mean trait values, 4) variance components and 5)
repeatabilities differed across nutritional environments
(experiment 2). We also tested 6) whether the effects of
diet (on mean and variance components) differed across
Cricket maintenance and diet preparation
For our experiments, we used the third generation of
offspring of adult crickets collected from a natural
population in Tuscany (Italy) in July 2013. By not using
offspring of wild-caught adults we thereby avoided biasing
effects of associated with environmental effects in the
wild [47, 48]. We worked with a laboratory setup as we
were interested in manipulating specific proximate factors
affecting behaviour under controlled conditions rather
than studying functional questions, which would warrant
field studies . We used both juveniles (3rd-4th instar)
and adults of the third laboratory generation for the two
experiments that we describe below. In experiment 1, we
measured (a) nutritional preferences of juveniles and
adults, and (b) nutritional intakes of juveniles and adults
forced on imbalanced diets (high-carbohydrate vs.
highprotein diets, see below). The latter question was
addressed by a sub-experiment assessing whether crickets
experience nutrient deficiency (or excess) under
imbalanced diets. In experiment 2, we subjected freshly emerged
nymphs of the 3rd laboratory generation to forced diet
treatments to test for the effects of diet on various
phenotypic traits. Stock and treatment individuals (eggs, nymphs
and adults) were all maintained at 26 °C with 60% relative
humidity under a 14 L:10D photoperiod. Nymphs and
adults were kept separately in groups of 30 individuals
housed in transparent plastic containers (23 × 15 × 17 cm).
Containers contained pieces of egg carton for shelter, a
plastic water bottle plugged with cotton wool, and a dry
bird food (Aleckwa Delikat, Germany) in the case of stock
We created artificial diets consisting of 40% cellulose
and 60% nutrient content by manipulating protein and
carbohydrate content following Ref. . Protein consisted
of a 3:1:1 mixture of casein, albumen, and peptone, and
carbohydrates of a 1:1 mixture of sucrose and dextrin. All
artificial diets contained Wesson’s salts (2.5%), ascorbic
acid (0.275%), cholesterol (0.55%) and a mix of vitamins
Experiment 1 – tests of diet preference
Our main study aim was to quantify differences in mean
and variance components between two alternative diet
treatments (detailed below). The interpretation of
associated results warrants information on i) the nutritional
preferences for macronutrients (e.g., protein and
carbohydrate) in the study species (experiment 1-1), and ii) intake
of macronutrients in the imbalanced diet treatments
(high-carbohydrate vs. high-protein diets, see below)
(experiment 1-2). Crickets can suffer a deficit of protein
or an excess of carbohydrate when faced with
highcarbohydrate imbalanced diet (see Fig. 1 in ) but
such information was not available for our study species.
We therefore designed an experiment consisting of two
sub-experiments, aimed at quantifying 1) nutritional
preferences and 2) nutritional intakes (of juveniles and adults)
of crickets faced with imbalanced diets.
Experiment 1-1. Nutritional preference test
To measure nutritional preferences for carbohydrate and
protein by juveniles, 10 juveniles (3rd-4th instar) were
collected from the stock population 2 weeks after
hatching. Each cricket was housed individually in a
transparent plastic container (23 × 15 × 17 cm) furbished with
pieces of egg carton as shelters, a plastic water bottle
plugged with cotton wool, and provided with two dishes.
The first dish contained an artificial diet (weighing about
500 mg) containing a 1:29 C:P ratio, while the second
dish contained the reverse ratio (29:1) of the same
weight. Because juveniles and adults subjected to these
nutritional preference tests were taken from our stock
populations, and thus previously fed on a dry bird food
(see above), we expected them to require some time to
adjust to the synthetic diet. We gave them (five to) seven
days of acclimation time to the novel diets before
collecting preference data. After the acclimation period, we
removed both dishes, weighed their dry mass to
determine food consumption, and replaced them with fresh
food. This procedure was subsequently repeated with a
4-day inter-test interval four times in total. Foods were
dried in a desiccating oven at 40 °C before introduction
and after removal; consumption rate was defined as the
difference between the before and after weight.
Similarly, to measure carbohydrate and protein intake
by adults, 37 freshly emerged adult crickets (19 males
and 18 females) were allocated to these two
complementary foods (two separate food dishes: 29:1 C:P and 1:29
C:P). Each cricket was placed in the same condition as
described above for the diet preference test of juveniles.
After a week of acclimation, dry mass of food consumed
(mass change in the food) was recorded 2 times with a
3-day inter-test interval.
Experiment 1-2. Nutritional intake test under imbalanced
To assess whether juveniles or adults experienced nutrient
deficiency or excess when faced with an imbalanced
dietary condition (see Fig. 1 in ), we measured the amount
of nutrients (carbohydrate and protein) consumed by an
individual in a high-carbohydrate (5:1 C:P) or high-protein
(1:5 C:P) diet. 12 juveniles, 5 adult males and 5 adult
females were allocated to the high-carbohydrate (5:1 C:P)
diet, whereas 12 further juveniles, 5 further adult males
and 5 further adult females were allocated to the
highprotein (5:1 C:P) diet. We used the same procedures to
measure the consumption rate as detailed above.
Experiment 2 – testing phenotypic effects of diet
This experiment was designed to test how juvenile and
adult diet affected phenotypes and their variance
components of both sexes. Using approaches detailed in the first
section of the methods, we created two different diets: a
high-protein (1:5 C:P) versus high-carbohydrate diet (5:1
C:P). We then provided the crickets with an experimental
diet using a two-way factorial design implemented as a
split-brood design, resulting in 4 groups (CC:
highcarbohydrate juvenile and high-carbohydrate adult diet;
CP: high-carbohydrate juvenile and high-protein adult
diet; PC: high-protein juvenile and high-carbohydrate
adult diet; PP: high-protein juvenile and high-protein adult
diet; Fig. 2). To do so, eggs were collected from our 3rd
generation of the stock population (see above). Once
nymphs from the stock population had hatched, they were
randomly assigned to a high-carbohydrate or a
highprotein juvenile diet, and raised as a group of 30
individuals in transparent plastic containers (23 × 15 × 17 cm)
(for animal husbandry conditions, see above). Upon
reaching their last instar, individuals were checked daily
to determine date of eclosion to adulthood. On the day
of eclosion, adults were randomly assigned to the
highcarbohydrate or the high-protein adult diet treatment.
Each adult individual was maintained in an individual
container with an egg carton for shelter and supplied
with water and the allocated diet ad libitum, which we
replaced every 3 days.
When individuals had received 4 weeks of adult diet
treatment, we performed a set of behavioural assays,
resulting in a total sample size of 109 treated males
(CC:17, CP:15, PC:30, PP:47), and 142 treated females
(CC:21, CP:27, PC:48, PP:46); we measured exploration,
aggression and mating activity. Prior to the initiation of
the behavioural assays, each male was identified with a
small dot of (Testors enamel) paint on its pronotum. On
the same day, exploration, aggression and mating activity
were measured in a fixed order with 1-3 minutes
between each test. The fixed order of the assays ensured
that all individuals experienced the exact same
conditions, which facilitates comparison between individuals
[52, 53]. Each individual was assayed for each of these 3
behaviours 4 times with a 2-day test-interval. Based on
published power analyses , this study design (i.e.,
number of repeats per individual) yielded high enough
statistical power to detect low levels of repeatability. All
the behavioural assays were recorded with a digital
camcorder and analysed with tracking software, Noldus
Ethovision XT 10 (Noldus Information Technology).
Fig. 2 Experimental design used to manipulate diets during juvenile
and adult stages: high-carbohydrate juvenile diet & high-carbohydrate
adult diet (CC), high-carbohydrate juvenile diet & high-protein adult
diet (CP), high-protein juvenile diet & high-carbohydrate adult diet (PC)
and high-protein juvenile diet & high-protein adult diet (PP)
All assays were performed on a rack fitted with 2 shelves,
each equipped with a camera, in the same climate room
where the individuals had been reared. Before the assays
were conducted, we randomly selected 4 males and 4
females, and randomly assigned 2 males and 2 females per
shelf. Each batch of 8 individuals was subsequently assayed
simultaneously (2 males and 2 females per shelf) in a fixed
order (exploration, aggression and mating activity). Within
a batch, all four individuals were first simultaneously
subjected to an exploration assay alone (Additional file 1:
Figure S2). Ten minutes following the onset of the
exploration assay, the 2 same-sex individuals of the same shelf
were placed together in a single arena to measure
aggressiveness (Additional file 1: Figure S2). Ten minutes
following the onset of the aggression assay, the 2 same-sex
individuals were separated, and the male and female of the
same shelf placed together in the same arena to assay
mating activity (Additional file 1: Figure S2). For the aggression
and mating behaviour tests, we randomly assigned
individuals into dyads (see above) without considering their diet
treatments. As each individual was repeatedly assayed but
not always with the same partner, our study design was
perfectly suited to estimate how much of the phenotypic
variance expressed by focal individuals (in aggression or mating
behaviour) was attributable to the identity of the focal
individual, the identity of its social partner, or residual variance
[34, 54–56]; partner identity effects estimated by our study
design represented the combined influences of indirect
genetic and indirect permanent environmental effects [57–59].
Fifteen minutes following the onset of the mating activity
assay, all individuals of the batch were returned to their
private rearing containers. Following the execution of all
assays within a batch, the testing arenas were thoroughly
cleaned and sand was exchanged for fresh sand to minimize
any possible effects of remaining contact pheromones.
Novel environment assays
Individuals were removed from their individual
containers and placed in a plastic arena with a removable
partition in the middle which created two small rooms
(15 × 15 × 10 cm, Additional file 1: Figure S2). The two
individuals were separated by an opaque partition in the
middle of the arena during the novel environment assay
(Additional file 1: Figure S2). In each compartment,
fine-grained white sand was spread on the bottom, and a
plastic semi-cylinder was provided as a shelter (Fig. 2 in
). The tracking software then measured each
individual’s total distance moved in the compartment for 10
minutes [55, 60]. The compartment was novel to the
crickets, and we essentially quantified how much time
individuals spend out of the shelter, exploring the novel
environment. We therefore labelled this behavioural
variable ‘exploration behaviour’ .
Aggression assays began when the shelters and the
partition in the middle of the arena were removed 10 minutes
after the onset of the exploration assay (Additional file 1:
Figure S2). Two same-sex individuals then interacted
and showed behaviour ranging from low-level aggression
(e.g., antennal fencing, threat postures) to high-level
aggression (e.g. aggressive song stridulation, flaring
mandibles and biting) . Each aggression test lasted for
10 minutes. The focal and opponent identities were
randomly assigned after the experiment for statistical
analyses. The tracking software measured the duration of
attack when the focal individual chased the opponent
(within 6 cm) to attack it; this measure represents an
appropriate proxy for aggressiveness based on our previous
work on this species .
Mating activity assays
Following the aggression assays (performed for males and
females simultaneously tough separately), we gently moved
both a male and a female into the same plastic arena
(15 × 15 × 10 cm) and left both to acclimatize for
30 seconds. The arena was of the same size as the one
used for the novel environment assay. During such tests,
once males recognized females, they typically start
courting females using courtship stridulation [63, 64]. Once a
female was attracted and mounted the courting male, the
male would normally attempt to transfer his
spermatophore. Following copulation, the male typically remains
close to the female to increase the duration of
spermatophore attachment  and attempts to remate [66, 67].
Therefore, how closely the male remained in the proximity
of the female indicates how actively the male courts
before, and guards after, copulation. We thus measured 1)
the average distance between the body centre of the male
and the body centre of the female as a proxy for male
mating activity, which we did for 15 minutes, and 2) female
latency to mount as a proxy for female mating activity.
Females who did not respond to males producing
courtship stridulation were given a mating activity (i.e., female
latency) score of 15 mins (26% of the total number of
female responses were scored as 15 mins). Notably, females
do not approach males that fail to produce courtship
stridulation; we therefore did not analyse female mating
activity data when the male did not produce courtship
song as this typically resulted in a failure of a female to
mount a male.
We weighed each individual to the nearest 0.001 g at the
end of the second and the fourth set of behavioural assays.
Individuals were euthanized by placing them in a -20 °C
freezer for 24 h following the last set of behavioural tests,
dried at approximately 50 °C for 48 h, and weighed. We
then extracted lipids from the dried crickets using two
24 h washes of chloroform , and samples were again
dried for 24 h and weighed. Lipid masses were calculated
by subtracting sample lean (lipid extracted) dry masses
from sample dry masses.
Experiment 1 – Tests of diet preference
We used unpaired t-tests to test for differences in 1)
carbohydrate intake, 2) protein intake and 3) total
nutrient intake (total consumption of carbohydrate and
protein) between treatments or sexes.
Experiment 2- Diet effects on mean phenotype
To calculate the effect of diet on an individual’s mean
phenotype (assessed for exploration, aggression, mating
activity and body weight), we used univariate
mixedeffects models including juvenile diet treatment
(highcarbohydrate diet as the contrast), adult diet treatment
(high-carbohydrate diet as the contrast), their interaction,
sex (females as the contrast), interactions between diet
treatments and sex, time of the day (hour), testing order,
and shelf (a 2-level factor, upper and lower shelves) as
fixed effects. Time of the day was mean-centred at the
population level. We also fitted individual identity as a
random effect; partner identity was also included as an
additional random effect in the analyses of aggression and
mating activity (following [34, 54]). Prior to analysis,
response values were standardised (mean = 0, SD = 1) to
ease interpretation. We assessed the significance of fixed
effects using Wald F-tests, and the significance of random
effects using likelihood ratio tests (LRTs). The test statistic
associated with the LRT was calculated as twice the
difference in log likelihood between models with vs. without a
random effect. To test the effect of the focal individual’s
identity or the partner individual’s identity, the value of
P was calculated using a mixture of P(χ2, df = 0) and
P(χ2, df = 1) [69–71].
Because we measured lipid mass of individuals once,
we did not acquire repeated measures for this trait. For
its analysis, we therefore used a general linear model,
where we included juvenile diet treatment, adult diet
treatment, their interaction, sex, interactions between
diet treatments and sex, and testing order as fixed
effects. We assessed the significance of fixed effects using
Additionally, since we randomly paired individuals in
the aggression and mating behaviour assays without
considering their diet treatments, we additionally included
the fixed effects of (i) the partner’s treatment and (ii) the
interaction between the focal and partner treatment on
the focal individual’s behaviour for analyses of
aggressiveness and mating behaviour.
Experiment 2- Diet effects on variance components and
To test for diet effects on among- and within-individual
variances, total phenotypic variances or repeatabilities,
we used multivariate mixed-effects models which
simultaneously fitted the same trait measured in the four
within-sex treatment groups (CC, CP, PC or PP (Fig. 2);
subscripts M vs. F used for males vs. females) as the four
response variables (e.g., EXPCC(M), EXPCP(M), EXPPC(M),
EXPPP(M) for the male dataset: exploration of males
reared under the CC/CP/PC/PP treatment). We fitted
covariates (time of the day or testing repeat) as fixed effects.
Depending on the response variable, we also included the
focal individual’s identity, or both the focal and partner
individual identities (for aggression and mating activity), as
In models built to test for diet effects on variance
components (i.e., among-individual variances (VI),
withinindividual variances (VR) or total phenotypic variances
(VP)), we used the untransformed trait values as response
variables. In models fitted to test diet effects on
repeatabilities, we used z-transformed trait values (i.e., mean = 0 and
SD = 1) instead. As repeatability is calculated as the VI
divided by VP, VI represents the repeatability for
ztransformed data (i.e., where VP = 1) . In the
multivariate mixed-effect model, due to the study’s design,
covariance parameters were not estimable and therefore
constrained to zero at all levels.
To test for the significance of diet effects on VI, VR, VP
or repeatability, we compared the multivariate model,
which fitted treatment-specific variances at the focal
identity level (detailed above; model 2-4, Table 1), with a
reduced (i.e., ‘null’) model in which the focal variance
component was constrained to be equal across the four
datasets (model 1 where VCC = VCP = VPC = VPP). To
explore whether juvenile diet affected variance
components or repeatabilities, we compared our null model
(model 1) with a model (model 2) where variances were
constrained to be equal for treatments sharing the same
juvenile diet (i.e., [VCC = VCP] ≠ [VPC = VPP]). Similarly,
to test whether adult diets affected variance components
or repeatabilities, we compared a null model (model 1)
with a model (model 3) where variances were constrained
to be equal for treatments sharing the same adult diet (i.e.,
[VCC = VPC] ≠ [VCP = VPP]). Finally, to test whether
juvenile and adult diet treatments were both important (either
due to additive or interactive effects), we compared a null
model (model 1) with a heterogeneous model where all
variances were free to vary (model 4). To test whether diet
treatment affected the total phenotypic variance, we
removed all random effects from the model and then
compared the residual (i.e., phenotypic) variance among
treatments as detailed above. We compared models
using likelihood ratio tests (LRTs). As an alternative
approach, we also compared models based on their Akaike
Information Criterion (AIC) weights (a measure of relative
support for each a priori considered model structure)
[72, 73]. AIC weight is the probability that the
candidate model would be the best fitting model among the
set of models considered. Because both approaches led
to the same conclusions, we used results from LRTs in
the main text (Table 3), and provide the analysis based
on AIC weights as supplementary material (Table S1).
All the statistical analyses were implemented in ASReml
3.0 and solved by restricted maximum likelihood.
Experiment 1 - Tests of diet preference
The ratio of average carbohydrate consumption to average
protein consumption (C:P) by juveniles over an entire week
(e.g. ) was 1:1.3. This finding implied that 2-week-old
juveniles preferred to eat relatively similar amounts of
carbohydrates (mean ± SE: 20 ± 3 mg) versus proteins
(15 ± 4 mg; Fig. 3a). Adults of both sexes, by contrast,
consumed relatively more carbohydrates: male diet
preference (i.e., C:P ratio) was 5.7:1, and female diet
preference was 2.7:1. When adults were able to choose
and consume both a high-carbohydrate and a high-protein
diet simultaneously, adult males consumed less
carbohydrates (t35 = -3.14, P = 0.003, Fig. 3b) and less protein
(t35 = -3.60, P < 0.001, Fig. 3b) than females. Altogether,
males therefore consumed less nutrients (calculated as the
summed total of carbohydrate and protein consumption)
compared to females (t35 = -3.54, P = 0.001).
Crickets of both age classes consumed less carbohydrates
and more protein compared to their preferred intake when
forced on an imbalanced high-protein (i.e., 1:5 C:P) diet
(Fig. 3a,b). Juveniles and adult females consumed less
protein but similar amounts of carbohydrates compared to
their preferred intake when forced on a high-carbohydrate
diet (5:1 C:P) (Fig. 3a,b). By contrast, adult males forced on
a high-carbohydrate diet (5:1 C:P), had intake rates of both
macronutrients that were close to their intake target (i.e.,
5.7:1 C:P, see above; Fig. 3b) as adult males generally prefer
relatively carbohydrate-biased diets.
Experiment 2 – diet effects on mean phenotype
Juvenile diet strongly affected morphological and
behavioural traits of both sexes other than exploration behaviour
and female mating activity (Table 2; Additional file 1: Figure
S3). Individuals raised on high-protein juvenile diets were
more aggressive and heavier as adults compared to
individuals raised on high-carbohydrate juvenile diets (Table 2;
Table 1 Overview of the series of multivariate mixed-effect models fitted and compared to estimate diet effects on individual
differentiation in behaviour and within-individual behavioural stability
VCC = VCP = VPC = VPP
[VCC = VCP] ≠ [VPC = VPP]
[VCC = VPC] ≠ [VCP = VPP]
Null model – Homogeneity of variance components across all diet treatments
Effect of juvenile diet - V was constrained the same within the same juvenile diet treatments
Effect of adult diet - V was constrained the same within the same adult diet treatments.
Model 4 (M4) VCC ≠ VCP ≠ VPC ≠ VPP Additive/non-additive effect of juvenile and adult diet -Unconstrained model.
Models differ in whether treatment-specific variance components (V) were estimated as distinct or constrained to be identical. This procedure was applied to study
treatment effects on either among-individual (VI) or within-individual variances (VR)
Additional file 1: Figure S3). Males raised on high-protein
juvenile diets also courted and guarded females more
closely compared to males raised on high-carbohydrate
juvenile diets (Table 2; Additional file 1: Figure S3). Given the
way that we set up the experiment (see Methods), this latter
finding implied either (i) that juvenile diet affected both
aggressiveness and mating behaviour, or (ii) that mating
behaviour was affected by carry-over effects associated
with aggressive individuals winning fights (see Discussion).
While juvenile diet treatment affected various
behaviours, this was not the case for adult diet treatment:
adult diet affected only body weight and lipid mass
(Table 2; Additional file 1: Figure S3). Individuals forced
on a high-protein adult diet gained more weight but had
lower lipid mass compared to those forced on a
highcarbohydrate adult diet (Table 2; Additional file 1: Figure
S3). Across diet treatments, males were more aggressive
and explorative in a novel environment than females
(Table 2). Females were heavier and their bodies contained
more lipids than males (Table 2). There was no evidence
for significant sex differences in the effects of juvenile or
adult diet treatment for any phenotypic trait except for
body weight and lipid mass (Table 2; Additional file 1:
Figure S3): Both sexes had more lipids in their body
when they were reared on the high-protein juvenile diet
but the difference in body lipids between juvenile diet
treatments was larger in females compared to males.
Similarly, both sexes gained more weight when they
were forced on the high-protein adult diet but the
difference in body weight between adult diet treatments
was larger in females compared to males.
For the interactive phenotypes (aggression and mating
activity), the partner’s diet treatment affected aggression
and mating behaviour in females (Additional file 1: Table
S3, female aggression, F4,138.7 = 3.10, P = 0.02; female mating
activity, F4,123.3 = 3.19, P = 0.02). Females tended to be more
aggressive and more active in the mating assay when they
encountered a partner that had received a high-protein
compared to a high-carbohydrate adult diet. However, the
effect of partner treatment did not affect the significance of
effects of other factors (Additional file 1: Table S3).
Experiment 2 – diet effects on variance components and
We estimated the among-individual, within-individual, and
total phenotypic variance for each unique combination of
juvenile and adult treatment and phenotypic trait (for age-,
treatment-and trait-specific estimates of each variance
component see Table S2 in the supplementary material). Diet
affected the expression of the total phenotypic,
amongand/or within-individual variances of various traits (Fig. 4;
Table 3). However, diet treatment did not simultaneously
affect among- and within-individual variances such that it
significantly changed any of the trait repeatabilities (Table 4).
Diet treatment affected the total phenotypic variance
for aggression, body weight and lipid mass in both sexes
(Fig. 4; Table 3). For aggression and lipid mass, the
highcarbohydrate adult diet increased the total phenotypic
variance when crickets were reared on a high-protein
juvenile diet (Fig. 4). This effect was not observed when
crickets were reared on a high-carbohydrate juvenile diet
(Fig. 4). In contrast, differences in the total phenotypic
variance in female body weight between adult diet
treatments were larger when juveniles were raised on
high-carbohydrate than on high-protein diets (Fig. 4).
Males raised on the high-protein juvenile diet
exhibited larger within-individual variance in aggression and
larger among-individual variance in weight compared to
males raised on the high-carbohydrate juvenile diet
(Fig. 4, Table 2). Females raised on the high-protein
juvenile diet exhibited larger within-individual variance in
body weight compared to females raised on the
highTable 2 Linear (mixed) models of behavioural and morphological traits as a function of diet, sex and their interaction. Parameters
are provided with standard errors in parentheses
ajuvenile diet effect (high-carbohydrate diet as the contrast)
badult diet effect (high-carbohydrate diet as the contrast)
csex effect (females as the contrast)
dinteractive effect between juvenile and adult diets
Significant terms are indicated in bold
carbohydrate juvenile diet (Fig. 4, Table 3). In addition,
the among-individual variance in male mating activity
differed among four diet combination treatments (Fig. 4,
Table 2): when males were reared on the high-carbohydrate
juvenile diet, the adult diet containing more carbohydrate
increased the among-individual variance in male mating
activity while adult diet did not affect the among-individual
variance in male mating activity when males were reared
on the high-protein juvenile diet (Fig. 4, Table 2).
There were no sex differences in the effect of diet
treatment on the total phenotypic variance (Table 3).
However, males and females differed in how diet affected
among- or within-individual variance components (Table 3).
While variance components in female aggression and
mating activity were not influenced by diet treatment,
the among-individual variance in male mating activity
and the within-individual variance in male aggression
were diet-dependent (Table 3).
Fig. 4 Variance components of phenotypes across four combinations of diet treatments (CC, high-carbohydrates juvenile & high-carbohydrates
adult diet treatment; CP, high-carbohydrates juvenile & high-protein adult diet treatment; PC, high-protein juvenile & high-carbohydrates adult
diet treatment; PP, high-protein juvenile & high-protein adult diet treatment). a male exploration, b female exploration, c male aggression,
d female aggression, e male mating activity, f female mating activity, g male body weight, h female body weight, (i) male lipid mass and
j female lipid mass. Stacked bars (y-axis, left) indicate the total phenotypic variance decomposed into its among-individual (dark grey bars),
residual within-individual (white bars) variance components and a variance explained by the interacting partner individual (black bars). Dots
(y-axis, right) represent level of repeatability (± s.e.)
One of the key meta-analytical findings in animal
personality research is that about 50% of the individual
differences in average behaviour are attributable additive
genetic effects . Our study shows that nutritional
history partly explains the remaining variation, causing
environmentally-underpinned repeatable differences in
behaviour. The nutritional environment affected both the
amount of individual differentiation and the amount of
within-individual stability in various phenotypic traits: four
of five traits assayed in males (including two of three
behaviours), and three of five traits assayed in females
(including one of three behaviours), showed diet-dependent
variance components (Table 3). Nutritional history during
ontogeny in particular represented an important
environmental factor mediating such non-genetic differences in
‘personality’: crickets raised on a high-protein diet
developed a more aggressive phenotype in adulthood compared
to those raised on a high-carbohydrate diet; males raised
on a high-protein diet also courted females more actively
Table 3 Diet effects on variance components (VI, among-individual variance; VR, within-individual variance; VP, total phenotypic variance)
and repeatability (R) of exploration, aggression, mating activity, weight and lipid mass)
Diet effect χ2df P χ2df
VI Juvenile 0.041 0.84 1.181
Adult 1.161 0.28 1.051
Combined 1.243 0.74 1.783
VR Juvenile 1.761 0.18 2.701
Adult 0.341 0.56 0.441
Combined 5.663 0.13 3.513
VP Juvenile 0.241 0.62 14.581
Adult 0.841 0.36 2.961
Combined 1.583 0.66 20.243
R Juvenile 0.111 0.74 0.401
Adult 0.981 0.32 0.921
Combined 1.133 0.77 1.243
χ2-values were derived from likelihood ratio tests (see Methods)
Significant effects are indicated in bold
in adulthood. Juvenile diet, furthermore, also affected an
individual’s level of behavioural stability: males raised on a
high-protein diet were not only more aggressive on
average but also behaviourally less stable (i.e., more variable or
‘changeable’) compared to males raised on a
carbohydraterich diet. Furthermore, we found some evidence for
interacting effects of early-life and adult diet on the individual
differentiation in behaviour, indicating the existence of
multidimensional plasticity [74, 75], though such
interactive effects were only detected for the individual
differentiation in male mating activity. Finally, for some traits,
patterns of diet-specific behavioural stability differed
between the sexes. The effect of diet on behavioural stability
and individual differentiation was relatively weak in females
compared to males.
Diet influences personality and morphology
Our study showed that the nutritional environment at the
juvenile stage affected the development of behavioural
(namely; aggression and mating activity) and
morphological traits (Table 2). Juveniles raised on low-protein and
high-carbohydrate (5:1 C:P) diets were not able to reach
their preferred intake of proteins (Fig. 3a). Proteins are a
limiting nutrient under natural field conditions [76, 77]
and strongly affect development . In insects, it is
typical that protein deficiency during development causes
Table 4 Within-treatment repeatabilities of behaviours and
body weight with standard errors in parentheses
R (SE) Exploration Aggression Mating activity
0.68 (0.04) 0.30 (0.09) 0.14 (0.05)
An average adjusted repeatability (Radj) for the whole dataset is presented,
and calculated after including diet treatment as a fixed effect factor into the
model. CC, high-carbs juvenile & high-carbs adult diet treatment; CP, high-carbs
juvenile & high-protein adult diet treatment; PC, high-protein juvenile & high-carbs
adult diet treatment; PP, high-protein juvenile & high-protein adult diet treatment
Variance components at the boundary are estimated as 0.00 with 0.00 SE
Significant values (P < 0.05) are indicated in bold
smaller body size and longer development time [79–81];
malnutrition during development is also known to reduce
juvenile and adult body lipid content [43, 82]. In
agreement with previous findings, individuals reared on
highprotein juvenile diets gained more weight and body lipids,
and individuals reared on high-carbohydrate juvenile diets
suffered increased mortality (CS Han & NJ Dingemanse,
unpublished data). This may explain why males on
highprotein juvenile diets were also more aggressive towards
conspecifics and more active in courtship and
postcopulation mate-guarding (Table 2). Therefore, the
behavioural effects of juvenile diet were likely mediated by the
acquisition of proteins required for development.
Adult diet also played a role in the expression of
morphology but not in the development of a cricket’s
personality. Adult diet was not able to offset the
detrimental effect of the juvenile high-carbohydrate diet
(‘silverspoon hypothesis’, ). When considering adult intake
targets (Fig. 3b), males and females clearly suffered
carbohydrate deficiency when raised on a high-protein
adult diet. In contrast, among animals raised on the
high-carbohydrate adult diet, only females suffered
proteindeficiency while males instead consumed enough proteins
to meet their preferred needs. Because our adult diet
treatments were not extreme, individuals confronted with a
high-protein adult diet could probably acquire energy also
by consuming more protein, which may explain why adult
diet generally did not elicit effects on the expression of
assayed phenotypic traits.
In addition, given the fixed sequential nature of our
experiments (where mating activity was always measured
after the aggression assay), we suggest caution in
interpreting these results: though diet affected both male
aggression and male mating activity, the outcome of the
contest might have directly spilled-over to affect mating
behaviour. Such carry-over effects of social interactions
are generally expected , in this particular case
possibly mediated by winner and/or loser effects (e.g.,
[85–87]). If so, we would expect a positive correlation
between aggression and mating activity, for example,
because less aggressive crickets more readily lose fights,
which in turn would inhibit subsequent mating activity. In
our experiment, male aggression indeed tended to be
positively correlated with male mating activity at the
among-individual level (bivariate mixed-effects model:
r = -0.40 ± 0.21, P = 0.08). This finding is consistent with
the notion that aggressive males had higher mating
activity because they won fights. However, this positive
correlation may also be due to a genetic correlation
between aggressiveness and mating behaviour that is not
mediated by winner-loser effects. Further experiments,
where testing order and inter-test intervals are
explicitly manipulated [53, 88, 89], are therefore required to
differentiate between these two alternative explanations.
Diet influences individual differentiation and behavioural
In this paper, we provide experimental evidence that
nutritional history affects the amount of individuality
(i.e., among-individual variance) and behavioural stability
(i.e., within-individual variance) of certain behavioural
traits. Detected changes in individual differentiation or
within-individual stability of phenotypes were related to
the juvenile rather than adult diet treatment (Table 2).
First, we showed that the high-protein juvenile diet
increased the mean level of aggression while simultaneously
decreasing its within-individual stability in adulthood
(though this latter effect existed only in males). This
finding thereby partly supports the hypothesis that both the
mean level and variance components (among- and
withinindividual variances) of phenotypic traits are greater in
high-quality (vs. low-quality) environments (Additional
file 1: Figure S1).
The following mechanisms may explain the simultaneous
increase in mean and level of instability in male aggression.
When two males interact, a behavioural hierarchy between
them is developed during the fight [90–92]. Following the
establishment of a hierarchy, the loser male rarely
approaches and attacks the winner male, whereas the winner
males actively chases and attacks the loser [90–92]. Males
developed in high-quality nutritional environments (i.e.,
our high-protein juvenile diet, Fig. 3b), where individuals
meet their protein needs, may gain more weight and thus
be more aggressive. This means that males in a good
condition can decide to express low levels of
aggressiveness when they lose fights, but they can act
hyperaggressively when they win the fight. This may explain
why high-quality nutritional environments can decrease
the within-individual stability but increase the mean
level of aggression.
Our study also showed that the nutritional environment
changes the amount of among-individual differentiation
and level of behavioural stability without affecting mean
levels of behaviour. For example, despite a combined
effect of adult and juvenile diets on the amount of individual
differentiation in male mating activity, the mean level of
male mating activity was only affected by juvenile diet
(Fig. 4). If some individuals increase their behavioural
level, but others decrease it, this may result in an increased
individual differentiation without changing mean values
(Additional file 1: Figure S1d). Since the expression of
behaviour is also mediated by trade-offs between life-history
traits , males facing nutritionally balanced dietary
environments (e.g. high-carbohydrate adult diets) may
vary their reproductive investment using abundant
resources. That is, some males may invest more in current
reproduction and increase the intensity of mating activity
while others instead decrease their current investment in
reproduction and mating activity. We thus suggest that
the extent of individual differentiation and behavioural
stability can change regardless of changes in mean value.
Taken together, our study shows that individual
differentiation in behaviour and behavioural stability were
independently responding to diet regardless of the effect on
the population-level mean (Additional file 1: Figure S1).
The effects of diet on personality and behavioural stability
were thus trait-specific (Table 3).
This study investigated the underlying proximate
mechanisms causing variation in among-individual
differentiation and within-individual behavioural stability
among datasets. Our findings imply that environmental
conditions such as diet can indeed explain the existence
of variation in these variance components, clarifying why
those may vary across datasets. High-quality environments
such as nutritionally balanced and pathogen-free
conditions may enable individuals in a good condition to
flexibly express costly behaviours in response to changes in
environments. An increase in resource availability (e.g.
protein) characterizing high quality environments may
also increase among-individual differences in how
tradeoffs (e.g., between lifespan and reproduction) are resolved,
thereby leading to further among-individual differentiation
in behaviour .
Our results are in line with previous research showing
that the expression of genetic variation is typically
increased under favourable conditions [19, 20], and recent
work by Royaute and Dochtermann  showing that
nutritional stress from low quality diet increased the
within-individual stability in the response to predators in
the house crickets (Acheta domesticus). By contrast,
some recent studies showed results opposite to ours: the
amount of individual differentiation in behaviour was
higher, and within-individual behavioural stability lower in
more stressful environments . Hermit crabs (Pagurus
bernhardus) also reduced their within-individual
behavioural stability in the presence of predators (i.e., a stressful
environment) because unpredictability in behaviour under
increased perceived predation risk might decrease their
actual predation risk . Recent work on zebra finches
(Taeniopygia guttata) also showed that early dietary
restriction increased the amount of among-individual
variance in activity . These lines of contrasting evidence
imply that the environment may directly affect the amount
of individual differentiation in behaviour and
withinindividual behavioural stability but that such effects might
often be trait-specific. Taken together, while our study
documents diet effects on variance components, the
contrasting evidence and trait-specific effects suggest
that further studies of this kind are required to assess
the generality of our findings across species.
Diet effects on repeatability
As repeatability is calculated as the proportion of the
among-individual variance relative to the total
phenotypic variance , a significant change in the amount of
among-individual variation, or level of behavioural
stability, is expected to cause a change in repeatability.
However, despite diet affecting the extent of individual
differentiation and level of behavioural stability, we failed
to find significant effects of diet on behavioural
repeatability (Table 3). Here we suggest explanations for
this finding. Firstly, diet may have little influence on
the repeatability when it affects a variance component
that accounts for a relatively small part of the total
phenotypic variance. For example, when the amount of
within-individual variance is much larger than the amount
of among-individual variance, only a large change in
among-individual variance might lead to a statistically
detectable change in repeatability across environments.
Alternatively, diet may affect repeatability only when it
influences the among- and within-individual variances
in different way (e.g. increases VI but decreases VR, as
in Ref. ). Overall the most plausible explanation is
that, while we had sufficient power to detect nonzero
repeatability within treatment groups, we simultaneously
lacked sufficient statistical power to detect significant
effects of the nutritional environment on repeatability.
Sex-specific diet effects
If males and females prefer different nutritional
environments, we may expect adaptive sex-differences in the
effects of diet. In our study, we found that females consumed
significantly more protein but less carbohydrate than males
when given the choice (i.e., we found evidence for
sexspecific nutritional preferences, Fig 3b). However, in
contrast to our prediction, there were no sex-specific effects
of diet on the mean level of behavioural traits. Females tend
to have high protein needs because proteins are required for
egg production . Females with access to much protein
during adulthood were thus able to invest more resources
(e.g., protein) in egg production, which resulted in a sharp
increase in their body weight. This may explain why
differences in body weight (and its associated within-individual
variance) were larger between adult diet treatments in
females compared to males. In addition, diet effects on
individual differentiation and behavioural plasticity were
relatively stronger for male compared to female behaviour
(Table 3). In our recent study, males were more vulnerable
to a deficit of protein than females (Han & Dingemanse:
Protein deficiency decreases male survival and the intensity
of sexual antagonism, submitted), suggesting that males
could be more sensitive to nutritional stress than females.
If this is the case, nutritional stress is likely to affect the
amount of individual differentiation and behavioural
plasticity more dramatically in males. However, sex-specific
effects on variance components might also be trait-specific.
Therefore further investigation of the relationship between
sex differences in nutritional preference and sex-specific
diet effects on variance components is required.
Animals often experience changes in nutritional
environments, and their nutritional requirements also change
throughout their lives . This study demonstrates the
importance of considering nutritional history as a proximate
explanation for variation in the extent of individual
differentiation and behavioural stability, thereby deepening our
understanding of the role of ecological factors shaping these
forms of variation. Furthermore, sex- and trait-specificity in
effects of nutritional history imply that a fruitful line of
future research would focus on the genetic architecture of
dietary preferences across sexes, which may be achieved by
using multivariate analyses and quantitative genetics
analyses . Finally, the integration of the concepts of
nutritional ecology and animal personality will provide a
major step towards a general understanding of behavioural
evolution in changing environments .
Additional file 1: Supplementary material. Figure S1. Schematic
representations of hypothesized population-level average behavioural
responses and associated among- and within-individual variances across
environments. Figure S2. The experimental set-up. Figure S3. The effect
of diet and sex on the expression of behavioural and morphological
traits. Table S1. Diet effects on variance components and repeatability of
phenotypes. Table S2. Variance components (with standard errors in
parentheses) for each unique combination of treatment for sex, and a
range of phenotypic traits. Table S3. Additional linear mixed models to
test the effect of social partner treatment on aggression and mating
activity. (DOCX 479 kb)
We thank Francesca Santostefano for the collection of crickets from the wild,
Yvonne Cämmerer and Bettina Rinjes for help in maintaining the stock
populations, and Robert Brooks, Michael Kasumovic, Ned Dochtermann and
Kwang Pum Lee for feedback and discussion. We also thank three anonymous
reviewers for their positive helpful comments.
CH and NJD designed the experiment. CH conducted experiments and
carried out the statistical analyses with input from NJD. CH and NJD wrote
the manuscript. All authors gave final approval for publication.
The authors declare that they have no competing interests.
Consent for publication
1. Dingemanse NJ , Kazem AJN , Réale D , Wright J. Behavioural reaction norms: animal personality meets individual plasticity . Trends Ecol Evol . 2010 ; 25 ( 2 ): 81 - 9 .
2. Dingemanse NJ , Wolf M. Recent models for adaptive personality differences: a review . Phil Trans R Soc B . 2010 ; 365 ( 1560 ): 3947 - 58 .
3. Westneat DF , Wright J , Dingemanse NJ . The biology hidden inside residual within‐individual phenotypic variation . Biol Rev . 2014 ; 90 ( 3 ): 729 - 43 .
4. Dochtermann NA , Schwab T , Sih A. The contribution of additive genetic variation to personality variation: heritability of personality . Proc R Soc Lond B Biol Sci . 2015 ; 282 ( 1798 ): 20142201 .
5. Kasumovic MM , Brooks RC . It's all who you know: the evolution of socially cued anticipatory plasticity as a mating strategy . Q Rev Biol . 2011 ; 86 ( 3 ): 181 - 97 .
6. Snell-Rood EC. An overview of the evolutionary causes and consequences of behavioural plasticity . Anim Behav . 2013 ; 85 ( 8 ): 1004 - 11 .
7. West-Eberhard MJ . Developmental plasticity and evolution . Oxford: Oxford University Press ; 2003 .
8. Meaney MJ . Maternal care, gene expression, and the transmission of individual differences in stress reactivity across generations . Annu Rev Neurosci . 2001 ; 24 ( 1 ): 1161 - 92 .
9. Weaver IC , Cervoni N , Champagne FA , D'Alessio AC , Sharma S , Seckl JR , Dymov S , Szyf M , Meaney MJ . Epigenetic programming by maternal behavior . Nat Neurosci . 2004 ; 7 ( 8 ): 847 - 54 .
10. Carere C , Drent PJ , Koolhaas JM , Groothuis TGG : Epigenetic effects on personality traits: early food provisioning and sibling competition . Behaviour , 142 2005 ; 9 ( 10 ): 1329 - 55 .
11. Dingemanse NJ , Dochtermann NA . Quantifying individual variation in behaviour: mixed-effect modelling approaches . J Anim Ecol . 2013 ; 82 ( 1 ): 39 - 54 .
12. Jenkins SH . Sex differences in repeatability of food-hoarding behaviour of kangaroo rats . Anim Behav . 2011 ; 81 ( 6 ): 1155 - 62 .
13. Briffa M. Plastic proteans: reduced predictability in the face of predation risk in hermit crabs . Biol Lett . 2013 ; 9 ( 5 ): 20130592 .
14. Careau V , Buttemer WA , Buchanan KL . Early-developmental stress, repeatability, and canalization in a suite of physiological and behavioral traits in female zebra finches . Integr Comp Biol . 2014 ; 54 ( 4 ): 539 - 54 .
15. DiRienzo N , Niemelä PT , Skog A , Vainikka A , Kortet R. Juvenile pathogen exposure affects the presence of personality in adult field crickets . Front Ecol Evol . 2015 ; 3 : 36 .
16. Royaute R , Dochtermann NA : When the mean no longer matters: Developmental diet affects behavioral variation but not population averages in the house cricket (Acheta domesticus) . Behav Ecol 2017 :in press.
17. Westneat DF , Schofield M , Wright J. Parental behavior exhibits amongindividual variance, plasticity, and heterogeneous residual variance . Behav Ecol . 2013 ; 24 ( 3 ): 598 - 604 .
18. Wood CW , Brodie ED. Environmental effects on the structure of the G‐matrix . Evolution . 2015 ; 69 ( 11 ): 2927 - 40 .
19. Charmantier A , Garant D. Environmental quality and evolutionary potential: lessons from wild populations . Proc R Soc Lond B . 2005 ; 272 ( 1571 ): 1415 - 25 .
20. Hoffmann AA , Merilä J. Heritable variation and evolution under favourable and unfavourable conditions . Trends Ecol Evol . 1999 ; 14 ( 3 ): 96 - 101 .
21. Han CS , Dingemanse NJ . Effect of diet on the structure of animal personality . Front Zool . 2015 ;12 Suppl 1: S5 .
22. Zajitschek F , Lailvaux SP , Dessmann J , Brooks R. Diet , sex, and death in field crickets . Ecol Evol . 2012 ; 2 ( 7 ): 1627 - 36 .
23. Blay S , Yuval B. Nutritional correlates of reproductive success of male Mediterranean fruit flies (Diptera: Tephritidae) . Anim Behav . 1997 ; 54 ( 1 ): 59 - 66 .
24. Droney DC . Environmental influences on male courtship and implications for female choice in a lekking HawaiianDrosophila . Anim Behav . 1996 ; 51 ( 4 ): 821 - 30 .
25. Hunt J , Brooks R , Jennions MD , Smith MJ , Bentsen CL , Bussiere LF . Highquality male field crickets invest heavily in sexual display but die young . Nature . 2004 ; 432 ( 7020 ): 1024 - 7 .
26. Maklakov AA , Simpson SJ , Zajitschek F , Hall MD , Dessmann J , Clissold F , Raubenheimer D , Bonduriansky R , Brooks RC . Sex-specific fitness effects of nutrient intake on reproduction and lifespan . Curr Biol . 2008 ; 18 ( 14 ): 1062 - 6 .
27. Wilder SM , Rypstra AL. Diet quality affects mating behaviour and egg production in a wolf spider . Anim Behav . 2008 ; 76 ( 2 ): 439 - 45 .
28. Lichtenstein JL , DiRienzo N , Knutson K , Kuo C , Zhao KC , Brittingham HA , Geary SE , Ministero S , Rice HK , David Z. Prolonged food restriction decreases body condition and reduces repeatability in personality traits in web-building spiders . Behav Ecol Sociobiol . 2016 ; 70 ( 11 ): 1793 - 803 .
29. Paaby AB , Rockman MV . Cryptic genetic variation: evolution's hidden substrate . Nat Rev Genet . 2014 ; 15 ( 4 ): 247 - 58 .
30. Senior A , Nakagawa S , Lihoreau M , Simpson S , Raubenheimer D. An Overlooked Consequence of Dietary Mixing: A Varied Diet Reduces Interindividual Variance in Fitness . Am Nat . 2015 ; 186 ( 5 ): 649 - 59 .
31. Bunning H , Bassett L , Clowser C , Rapkin J , Jensen K , House CM , Archer CR , Hunt J. Dietary choice for a balanced nutrient intake increases the mean and reduces the variance in the reproductive performance of male and female cockroaches . Ecol Evol . 2016 ; 6 ( 14 ): 4711 - 30 .
32. DeWitt TJ , Sih A , Wilson DS . Costs and limits of phenotypic plasticity . Trends Ecol Evol . 1998 ; 13 ( 2 ): 77 - 81 .
33. Wilson A , Morrissey M , Adams M , Walling C , Guinness F , Pemberton J , CluttonBrock T , Kruuk L. Indirect genetics effects and evolutionary constraint: an analysis of social dominance in red deer . Cervus elaphus J Evol Biol . 2011 ; 24 ( 4 ): 772 - 83 .
34. Wilson AJ , Gelin U , Perron M-C , Réale D. Indirect genetic effects and the evolution of aggression in a vertebrate system . Proc R Soc Lond B Biol Sci . 2009 ; 276 ( 1656 ): 533 - 41 .
35. Houle D. Comparing evolvability and variability of quantitative traits . Genetics . 1992 ; 130 ( 1 ): 195 - 204 .
36. Lee KP . Sex-specific differences in nutrient regulation in a capital breeding caterpillar , Spodoptera litura (Fabricius). J Insect Physiol . 2010 ; 56 ( 11 ): 1685 - 95 .
37. Lee KP , Kim J-S , Min K-J . Sexual dimorphism in nutrient intake and life span is mediated by mating in Drosophila melanogaster . Anim Behav . 2013 ; 86 ( 5 ): 987 - 92 .
38. Morehouse NI , Nakazawa T , Booher CM , Jeyasingh PD , Hall MD . Sex in a material world: why the study of sexual reproduction and sex‐specific traits should become more nutritionally‐explicit . Oikos . 2010 ; 119 ( 5 ): 766 - 78 .
39. Harrison SJ , Raubenheimer D , Simpson SJ , Godin J-GJ , Bertram SM . Towards a synthesis of frameworks in nutritional ecology: interacting effects of protein, carbohydrate and phosphorus on field cricket fitness . Proc R Soc Lond B Biol Sci . 2014 ; 281 ( 1792 ): 20140539 .
40. Simpson SJ , Raubenheimer D : The nature of nutrition: a unifying framework from animal adaptation to human obesity : Princeton: Princeton University Press ; 2012
41. Adler MI , Cassidy EJ , Fricke C , Bonduriansky R. The lifespan-reproduction trade-off under dietary restriction is sex-specific and context-dependent . Exp Gerontol . 2013 ; 48 ( 6 ): 539 - 48 .
42. Bauerfeind SS , Fischer K. Effects of food stress and density in different life stages on reproduction in a butterfly . Oikos . 2005 ; 111 ( 3 ): 514 - 24 .
43. Dmitriew C , Rowe L. The effects of larval nutrition on reproductive performance in a food-limited adult environment . PLoS One . 2011 ; 6 ( 3 ): e17399 .
44. Zajitschek F , Hunt J , Jennions MD , Hall MD , Brooks RC . Effects of juvenile and adult diet on ageing and reproductive effort of male and female black field crickets . Teleogryllus commodus Funct Ecol . 2009 ; 23 ( 3 ): 602 - 11 .
45. Maklakov AA , Hall MD , Simpson SJ , Dessmann J , Clissold FJ , Zajitschek F , Lailvaux SP , Raubenheimer D , Bonduriansky R , Brooks RC . Sex differences in nutrient-dependent reproductive ageing . Aging Cell . 2009 ; 8 ( 3 ): 324 - 30 .
46. Kasumovic MM , Hall MD , Try H , Brooks RC. Socially cued developmental plasticity affects condition-dependent trait expression . Behav Ecol . 2013 ; 24 ( 2 ): 429 - 34 .
47. Matos M. Maternal effects can inflate rate of adaptation to captivity . Proc Natl Acad Sci . 2012 ; 109 ( 36 ): E2380 .
48. Wolf JB , Wade MJ . What are maternal effects (and what are they not) ? Phil Trans R Soc B . 2009 ; 364 ( 1520 ): 1107 - 15 .
49. Niemelä PT , Dingemanse NJ . Artificial environments and the study of 'adaptive'personalities . Trends Ecol Evol . 2014 ; 29 ( 5 ): 245 - 7 .
50. Simpson S , Abisgold J. Compensation by locusts for changes in dietary nutrients: behavioural mechanisms . Physiol Entomol . 1985 ; 10 ( 4 ): 443 - 52 .
51. Raubenheimer D , Simpson SJ . Integrative models of nutrient balancing: application to insects and vertebrates . Nutr Res Rev . 1997 ; 10 : 151 - 79 .
52. Dingemanse NJ , Wright J , Kazem AJN , Thomas DK , Hickling R , Dawnay N. Behavioural syndromes differ predictably between 12 populations of three-spined stickleback . J Anim Ecol . 2007 ; 76 ( 6 ): 1128 - 38 .
53. Dochtermann N. Behavioral syndromes: carryover effects, false discovery rates, and a priori hypotheses . Behav Ecol . 2010 ; 21 ( 5 ): 919 - 26 .
54. Dingemanse NJ , Araya-Ajoy YG . Interacting personalities: behavioural ecology meets quantitative genetics . Trends Ecol Evol . 2015 ; 30 ( 2 ): 88 - 97 .
55. Santostefano F , Wilson AJ , Araya-Ajoy YG , Dingemanse NJ . Interacting with the enemy: indirect effects of personality on conspecific aggression in crickets . Behav Ecol . 2016 ; 27 : 1235 - 46 .
56. Wilson AJ , Grimmer A , Rosenthal GG . Causes and consequences of contest outcome: aggressiveness, dominance and growth in the sheepshead swordtail, Xiphophorus birchmanni . Behav Ecol Sociobiol . 2013 ; 67 ( 7 ): 1151 - 61 .
57. McGlothlin JW , Moore AJ , Wolf JB , Brodie III ED . Interacting phenotypes and the evolutionary process . III Social evolution. Evolution . 2010 ; 64 ( 9 ): 2558 - 74 .
58. Wolf JB , Brodie III ED , Cheverud JM , Moore AJ , Wade MJ . Evolutionary consequences of indirect genetic effects . Trends Ecol Evol . 1998 ; 13 ( 2 ): 64 - 9 .
59. Wolf JB , Brodie III ED , Moore AJ . Interacting phenotypes and the evolutionary process . II. Selection resulting from social interactions . Am Nat . 1999 ; 153 ( 3 ): 254 - 66 .
60. Han CS , Jäger H , Dingemanse NJ. Individuality in nutritional preferences: a multi-level approach in field crickets . Sci Rep . 2016 ; 6 : 29071 .
61. Réale D , Reader SM , Sol D , McDougall PT , Dingemanse NJ . Integrating animal temperament within ecology and evolution . Biol Rev . 2007 ; 82 ( 2 ): 291 - 318 .
62. Adamo S , Linn C , Hoy R. The role of neurohormonal octopamine during 'fight or flight' behaviour in the field cricket Gryllus bimaculatus . J Exp Biol . 1995 ; 198 ( 8 ): 1691 - 700 .
63. Simmons L. Female choice in the field cricket Gryllus bimaculatus (De Geer) . Anim Behav . 1986 ; 34 ( 5 ): 1463 - 70 .
64. Alexander RD , Otte D : The evolution of genitalia and mating behavior in crickets (Gryllidae): Misc Publ Mus Zool Univ Mich 1967 ; 133 : 1 - 62 .
65. Simmons L. Sperm competition as a mechanism of female choice in the field cricket, Gryllus bimaculatus . Behav Ecol Sociobiol . 1987 ; 21 ( 3 ): 197 - 202 .
66. Wynn H , Vahed K. Male Gryllus bimaculatus guard females to delay them from mating with rival males and to obtain repeated copulations . J Insect Behav . 2004 ; 17 ( 1 ): 53 - 66 .
67. Sakaluk SK . Post-copulatory mate guarding in decorated crickets . Anim Behav . 1991 ; 41 ( 2 ): 207 - 16 .
68. Loveridge J. Age and the changes in water and fat content of adult laboratory-reared Locusta migratoria migratorioides R. and F. Rhodesian J Agric Res . 1973 ; 11 : 131 - 44 .
69. Pinheiro J , Bates D. Mixed-effects models in S and S-PLUS. New York : Springer; 2006 .
70. Self SG , Liang K-Y . Asymptotic properties of maximum likelihood estimators and likelihood ratio tests under nonstandard conditions . J Am Stat Assoc . 1987 ; 82 ( 398 ): 605 - 10 .
71. Visscher PM . A note on the asymptotic distribution of likelihood ratio tests to test variance components . Twin Res Hum Genet . 2006 ; 9 ( 04 ): 490 - 5 .
72. Akaike H. Maximum likelihood identification of Gaussian autoregressive moving average models . Biometrika . 1973 ; 60 ( 2 ): 255 - 65 .
73. Burnham K , Anderson D : Model selection and inference. 2nd edition . New York : Springer; 1998 .
74. Westneat DF , Hatch MI , Wetzel DP , Ensminger AL. Individual variation in parental care reaction norms: integration of personality and plasticity . Am Nat . 2011 ; 178 ( 5 ): 652 - 67 .
75. Westneat DF , Stewart IR , Hatch MI . Complex interactions among temporal variables affect the plasticity of clutch size in a multi-brooded bird . Ecology . 2009 ; 90 ( 5 ): 1162 - 74 .
76. McNeill S. The dynamics of a population of Leptopterna dolabrata (Heteroptera: Miridae) in relation to its food resources . J Anim Ecol . 1973 ; 42 : 495 - 507 .
77. Iversen T. Ingestion and growth in Sericostoma personatum (Trichoptera) in relation to the nitrogen content of ingested leaves . Oikos . 1974 ; 25 : 278 - 82 .
78. Merkel G . The effects of temperature and food quality on the larval development of Gryllus bimaculatus (Orthoptera , Gryllidae). Oecologia . 1977 ; 30 ( 2 ): 129 - 40 .
79. Gebhardt MD , Stearns SC . Reaction norms for developmental time and weight at eclosion in Drosophila mercatorum . J Evol Biol . 1988 ; 1 ( 4 ): 335 - 54 .
80. Tu MP , Tatar M. Juvenile diet restriction and the aging and reproduction of adult Drosophila melanogaster . Aging Cell . 2003 ; 2 ( 6 ): 327 - 33 .
81. Colasurdo N , Gélinas Y , Despland E. Larval nutrition affects life history traits in a capital breeding moth . J Exp Biol . 2009 ; 212 ( 12 ): 1794 - 800 .
82. Stoks R , Block MD , McPeek MA . Physiological costs of compensatory growth in a damselfly . Ecology . 2006 ; 87 ( 6 ): 1566 - 74 .
83. Monaghan P. Early growth conditions, phenotypic development and environmental change . Philos Trans R Soc Lond B . 2008 ; 363 ( 1497 ): 1635 - 45 .
84. Niemelä PT , Santostefano F. Social carry-over effects on non-social behavioral variation: mechanisms and consequences . Front Ecol Evol . 2015 ; 3 : 49 .
85. Hsu Y , Earley RL , Wolf LL . Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes . Biol Rev . 2006 ; 81 ( 1 ): 33 - 74 .
86. Hsu Y , Wolf LL . The winner and loser effect: integrating multiple experiences . Anim Behav . 1999 ; 57 ( 4 ): 903 - 10 .
87. Kasumovic MM , Elias DO , Sivalinghem S , Mason AC , Andrade MC : Examination of prior contest experience and the retention of winner and loser effects . Behav Ecol . 2010 ; 21 ( 2 ): 404 - 409 .
88. Diaz-Uriarte R. Incorrect analysis of crossover trials in animal behaviour research . Anim Behav . 2002 ; 63 : 815 - 22 .
89. Senn S. Cross-over trials in clinical research , vol. 5. Chichester: Wiley; 1993 .
90. Adamo SA , Hoy RR . Agonistic behaviour in male and female field crickets, Gryllus bimaculatus, and how behavioural context influences its expression . Anim Behav . 1995 ; 49 ( 6 ): 1491 - 501 .
91. Sakura M , Aonuma H. Aggressive behavior in the antennectomized male cricket Gryllus bimaculatus . J Exp Biol . 2013 ; 216 ( 12 ): 2221 - 8 .
92. Hofmann HA , Stevenson PA. Flight restores fight in crickets . Nature . 2000 ; 403 ( 6770 ): 613 .
93. Stearns SC . The evolution of life histories . New York : Oxford University Press ; 1992 .
94. Nakagawa S , Schielzeth H. Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists . Biol Rev . 2010 ; 85 ( 4 ): 935 - 56 .
95. Wyman M , Stinchcombe J , Rowe L. A multivariate view of the evolution of sexual dimorphism . J Evol Biol . 2013 ; 26 ( 10 ): 2070 - 80 .