ArticlePDF Available

Abstract and Figures

Random and subtle deviations from bilateral symmetry (fluctuating asymmetry) have long been of interest to biologists who wish to study the susceptibility of organisms to changes in environmental quality. However, the reliability of FA as a biomarker has come under question due to inconsistent results in the literature. We conducted a meta-analysis of published literature to test the hypothesis that FA is a reliable biomarker of environmental stress in insects and identify possible sources of variation amongst studies. We expected studies to detect larger, positive magnitudes of effect on FA in lab populations due to the lack of confounding effects from other environmental factors compared to wild populations. Additionally, we predicted that studies that used geometric morphometric approaches to FA in shape and size would be more sensitive to changes in environmental quality compared to linear and meristic measures and thus show larger effects on FA. We also expected anthropogenic stressors to generate significantly larger effects on FA compared to naturally occurring stressors due to the organisms’ inability to buffer developmental pathways against a novel stressor. Finally, we predicted comparatively larger magnitudes of effect in studies that verified the environmental factor acting on the organism was a stressor by detecting negative effects on fitness-related traits. Overall, we found that FA is a sensitive biomarker of environmental stress. Environmental stressors explained 36% of the variation of effect on FA across studies. Studies that demonstrated a negative effect of the stressor on fitness-related traits showed significantly larger, positive magnitudes of effect on FA compared to studies that did not detect an effect from the environmental stressor. Additionally, studies conducted under laboratory conditions detected significantly larger, effects on FA compared to field-based studies. The kind of trait measured and the novelty of the stressor did not significantly account for differences amongst studies. Thus, the use of FA as a biomarker of environmental stress is a legitimate tool particularly when studies verify the biological relevance of stressors for the study organism.
Content may be subject to copyright.
This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/authorsrights
Author's personal copy
Ecological
Indicators
30
(2013)
218–226
Contents
lists
available
at
SciVerse
ScienceDirect
Ecological
Indicators
jo
ur
n
al
homep
ag
e:
www.elsevier.com/locate/ecolind
The
use
of
fluctuating
asymmetry
as
a
measure
of
environmentally
induced
developmental
instability:
A
meta-analysis
De
Anna
E.
Beasley,
Andrea
Bonisoli-Alquati,
Timothy
A.
Mousseau
University
of
South
Carolina,
Department
of
Biological
Sciences,
Columbia,
SC
29208,
USA
a
r
t
i
c
l
e
i
n
f
o
Article
history:
Received
22
September
2012
Received
in
revised
form
13
February
2013
Accepted
21
February
2013
Keywords:
Insects
Environmental
stress
Biomarker
Fluctuating
asymmetry
(FA)
Fitness-related
traits
Meta-analysis
Developmental
instability
a
b
s
t
r
a
c
t
Random
and
subtle
deviations
from
bilateral
symmetry
(fluctuating
asymmetry)
have
long
been
of
inter-
est
to
biologists
who
wish
to
study
the
susceptibility
of
organisms
to
changes
in
environmental
quality.
However,
the
reliability
of
FA
as
a
biomarker
has
come
under
question
due
to
inconsistent
results
in
the
literature.
We
conducted
a
meta-analysis
of
published
literature
to
test
the
hypothesis
that
FA
is
a
reliable
biomarker
of
environmental
stress
in
insects
and
identify
possible
sources
of
variation
amongst
studies.
We
expected
studies
to
detect
larger,
positive
magnitudes
of
effect
on
FA
in
lab
populations
due
to
the
lack
of
confounding
effects
from
other
environmental
factors
compared
to
wild
populations.
Additionally,
we
predicted
that
studies
that
used
geometric
morphometric
approaches
to
FA
in
shape
and
size
would
be
more
sensitive
to
changes
in
environmental
quality
compared
to
linear
and
meristic
measures
and
thus
show
larger
effects
on
FA.
We
also
expected
anthropogenic
stressors
to
generate
sig-
nificantly
larger
effects
on
FA
compared
to
naturally
occurring
stressors
due
to
the
organisms’
inability
to
buffer
developmental
pathways
against
a
novel
stressor.
Finally,
we
predicted
comparatively
larger
mag-
nitudes
of
effect
in
studies
that
verified
the
environmental
factor
acting
on
the
organism
was
a
stressor
by
detecting
negative
effects
on
fitness-related
traits.
Overall,
we
found
that
FA
is
a
sensitive
biomarker
of
environmental
stress.
Environmental
stressors
explained
36%
of
the
variation
of
effect
on
FA
across
studies.
Studies
that
demonstrated
a
negative
effect
of
the
stressor
on
fitness-related
traits
showed
sig-
nificantly
larger,
positive
magnitudes
of
effect
on
FA
compared
to
studies
that
did
not
detect
an
effect
from
the
environmental
stressor.
Additionally,
studies
conducted
under
laboratory
conditions
detected
significantly
larger,
effects
on
FA
compared
to
field-based
studies.
The
kind
of
trait
measured
and
the
novelty
of
the
stressor
did
not
significantly
account
for
differences
amongst
studies.
Thus,
the
use
of
FA
as
a
biomarker
of
environmental
stress
is
a
legitimate
tool
particularly
when
studies
verify
the
biological
relevance
of
stressors
for
the
study
organism.
©
2013
Elsevier
Ltd.
All
rights
reserved.
1.
Introduction
Monitoring
the
impacts
of
environmental
stressors
on
biological
systems
is
of
interest
to
biologists
concerned
with
management
and
sustainability
(Depledge
and
Galloway,
2005).
Biomarkers
are
defined
as
functional
measures
of
exposure
to
various
stressors
and
can
serve
as
an
early
warning
system
of
declines
in
environmental
quality
and
population
health
(Adams
et
al.,
2001).
They
play
a
role
in
determining
the
presence
of
a
stressor
(Leung
et
al.,
2003)
and
assessing
the
degree
to
which
important
taxonomic
groups
have
been
compromised
(Schlenk,
1999).
Fluctuating
asymmetry
(FA)
has
become
a
popular
method
for
measuring
phenotypic
response
to
environmental
stress
(Leamy
Corresponding
author.
Tel.:
+1
803
777
8241.
E-mail
addresses:
beasleyd@sc.edu,
deannaebeasley@gmail.com (D.E.
Beasley).
and
Klingenberg,
2005).
FA
is
defined
as
small,
random
deviations
of
bilateral
traits
from
perfect
symmetry
due
to
subtle
variations
in
the
developmental
environment
(Palmer
and
Strobeck,
2003).
Thus,
significantly
increased
levels
of
FA
in
a
population
may
indicate
that
individuals
are
having
more
difficulty
maintaining
precise
devel-
opment,
resulting
in
negative
effects
on
the
population
over
time
(Markow,
1995).
The
attractiveness
of
FA
as
a
potential
biomarker
stems
from
its
non-lethal
and
broad
application
across
biological
systems,
stressors
and
its
expected
association
with
life
history
traits
and
fitness
(Lens
et
al.,
2002;
Lens
and
Eggermont,
2008).
An
additional
advantage
of
FA
as
a
biomarker
includes
the
rela-
tive
ease
in
taking
measurements
compared
to
other
biomarkers
that
require
more
costly
equipment
(Leung
et
al.,
2003).
Given
the
difficulty
in
acquiring
direct
fitness
measures
from
populations
in
the
field,
and
the
importance
of
physiological
and
developmen-
tal
homeostases
for
fitness,
researchers
have
proposed
that
FA
serve
as
a
surrogate
for
estimating
how
well
a
population
is
coping
with
changes
in
environmental
conditions
(Clarke,
1995;
Møller,
1470-160X/$
see
front
matter
©
2013
Elsevier
Ltd.
All
rights
reserved.
http://dx.doi.org/10.1016/j.ecolind.2013.02.024
Author's personal copy
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
219
1997,
1999;
Møller
and
Thornhill,
1997,
1998).
However,
results
in
the
literature
have
been
inconsistent
across
species,
traits,
and
stressors,
causing
some
to
question
FA’s
reliability
as
a
biomarker
(Clarke,
1998;
Floate
and
Coghlin,
2010;
Floate
and
Fox,
2000;
Palmer,
1996).
For
example,
Eeva
et
al.
(2000)
found
increased
levels
of
FA
in
the
tarsus
of
great
tits
(Parus
major)
populations
exposed
to
heavy
metal
contamination.
However,
a
later
study
using
the
same
species
did
not
find
an
association
between
FA
and
heavy
metal
exposure
(Dauwe
et
al.,
2006).
A
similar
contradic-
tion
arose
when
Valkama
and
Kozlov
(2001)
failed
to
detect
an
increase
in
FA
levels
in
birch
(Betula
pubescens)
leaves
in
response
to
air
pollution,
as
shown
by
a
previous
study
(Kozlov
et
al.,
1996).
These
inconsistent
results
may
be
due
to
a
misuse
of
FA
with
respect
to
the
study
system.
For
instance,
the
setting
of
the
study
may
influence
the
relationship
between
FA
and
the
stressor.
In
field
studies,
organisms
are
exposed
to
many
differ-
ent
environmental
factors
that
may
reduce
the
strength
of
the
relationship
between
FA
and
the
stressor
in
question.
Similarly,
different
measures
of
FA
may
provide
more
or
less
detail
regard-
ing
the
morphology
of
the
trait
and
its
response
to
environmental
stress.
Studies
that
apply
more
sensitive
methods,
such
as
the
use
of
geometric
morphometrics
for
analyzing
biological
shape
and
size,
may
detect
a
larger
signal
compared
to
studies
that
use
linear
and/or
meristic
measures.
Furthermore,
while
the
general
definition
of
a
stressor
encompasses
any
biotic
or
abiotic
fac-
tor
that
interferes
with
an
individual’s
energy
allocation
toward
its
reproduction
and
development,
novel,
anthropogenic
stress-
ors
may
have
stronger
effects
on
FA
levels
because
they
more
actively
interfere
with
developmental
pathways
(Parsons,
2005)
and
more
directly
limit
the
mass
and
energy
available
compared
to
naturally
occurring
stressors,
against
which
adaptive
responses
are
expected
to
have
evolved
(Graham
et
al.,
2010;
Hoffmann
and
Hercus,
2000).
Additionally,
if
an
environmental
factor
is
not
significantly
interfering
with
the
expression
of
fitness-related
traits
such
as
body
size,
the
ability
to
complete
development
or
developmental
timing,
it
may
not
be
a
stressor
at
its
cur-
rent
level.
Thus,
we
would
not
expect
FA
to
significantly
vary
in
response.
Because
of
the
increasing
demand
for
simple
biomarkers
in
conservation
and
monitoring
studies
(Forbes
et
al.,
2006)
and
the
potential
use
of
FA
as
a
general
biomarker,
we
used
a
meta-
analytical
approach
to
assess
if
FA
is
a
reliable
biomarker
of
environmental
stress.
We
also
aimed
to
characterize
the
attributes
of
the
different
studies
that
explain
their
variation
in
the
strength
of
the
relationship
between
FA
and
environmental
stress.
We
expect
this
analysis
to
identify
the
limitations
of
the
application
of
FA
as
a
reliable
biomarker
and
provide
directions
concerning
the
study
design
and
techniques
that
are
more
likely
to
detect
a
biological
response
of
FA
to
a
stressor.
Meta-analyses
provide
a
systematic,
detailed
approach
for
identifying
findings
in
common
amongst
studies,
accounting
for
differences
in
study
design
and
integrating
results
in
a
coher-
ent
manner
(Arnqvist
and
Wooster,
1995;
Gurevitch
et
al.,
2001;
Stewart,
2010).
Additionally,
by
combining
the
results
of
multi-
ple
studies,
they
better
identify
the
true
magnitude
of
a
small
and
weak
relationship
between
two
factors
(Harrison,
2011).
Thus,
we
collected
a
sample
of
studies
that
applied
FA
as
a
biomarker
of
environmental
stress
and
determined
the
overall
magnitude
and
direction
of
effect
of
stress
on
FA.
We
grouped
studies
by
whether
they
conducted
the
experiment
in
the
field
or
under
laboratory
conditions
to
determine
if
the
degree
of
control
for
external
factors
outside
of
the
experimental
parameters
affected
the
reliability
of
FA.
Laboratory-based
studies
may
be
expected
to
detect
larger
effects
due
to
the
absence
of
confounding
effects
that
could
potentially
dilute
the
strength
of
the
relationship.
Next,
we
grouped
studies
by
the
type
of
environmental
stressor.
We
expected
anthropogenic
stressors
such
as
heavy
metal
contam-
ination
and
pesticides
to
have
significantly
larger
and
positive
magnitudes
of
effect
on
FA
compared
to
naturally
occurring
stress-
ors
such
as
changes
in
temperature.
In
naturally
occurring
stressors,
organisms
may
be
more
developmentally
resilient
due
to
evolved
adaptation
compared
to
organisms
exposed
to
novel
stressors
that
interfere
with
an
organism’s
ability
to
repair
and
maintain
develop-
mental
pathways
(Hoffmann
and
Hercus,
2000).
We
also
grouped
studies
by
the
method
in
which
they
measured
traits.
We
pre-
dicted
that
measurements
that
provide
more
detailed
information
about
the
morphology
of
the
trait,
such
as
geometric
morphometric
approaches
to
FA
in
shape
and
size,
will
be
more
sensitive
to
sub-
tle
changes
in
FA
compared
to
meristic
counts
or
linear
measures.
Finally,
we
categorized
studies
by
whether
or
not
they
detected
a
negative
fitness
consequence
via
measures
of
fitness-related
traits.
We
predicted
that
studies
with
evidence
of
a
fitness
consequence
due
to
exposure
to
the
environmental
factor,
verifying
that
the
factor
is
indeed
a
stressor,
would
result
in
larger,
more
positive
magnitudes
of
effect.
2.
Materials
and
methods
2.1.
Extracting
data
We
searched
Web
of
Science
(Web
of
Knowledge,
Philadelphia,
PA)
and
Google
Scholar
(Mountain
View,
CA)
for
articles
using
a
minimum
combination
of
two
of
the
following
keywords:
‘fluc-
tuating
asymmetry’
and
‘environmental
stress*’,
‘environmental
disturbance*’
and
‘fluctuating
asymmetry’,
‘environmental
stress
and
‘insect’.
We
also
searched
the
reference
lists
of
the
selected
articles
for
additional
studies
that
met
our
inclusion
criteria.
Our
meta-analysis
includes
studies
that:
(1)
used
insect
species,
(2)
addressed
the
question:
‘does
a
certain
environmental
stressor
affect
FA?’,
and
(3)
applied
FA
as
a
biomarker
of
the
impact
of
the
stressor
on
the
population.
We
limited
the
meta-analysis
to
insect
studies
because
it
is
important
to
use
a
homogenous
group
of
studies
to
better
draw
meaningful
conclusions
from
the
analysis
(Hunter
and
Schmidt,
2004).
We
felt
insects
were
an
appropriate
group
to
study
because
they
have
historically
been
used
to
assess
the
impacts
of
a
variety
of
stressors
on
environmental
quality.
Insects
have
many
characters
such
as
rapid
generation
turnover,
large
samples
and
sensitivity
to
environmental
changes
that
make
them
suitable
bioindicators
of
environmental
quality
(McGeoch,
1998,
2007).
We
included
the
outcome
for
all
traits
measured
in
each
study,
as
long
as
the
trait
met
the
statistical
assumptions
for
FA
analysis,
which
included
absence
of
directional
asymme-
try,
antisymmetry
and
trait-size
dependence
(Palmer
and
Strobeck,
2003).
Thus,
some
studies
contributed
data
for
more
than
one
effect
size
estimate.
We
excluded
studies
that
examined
the
relationship
between
FA
and
mating
success
because
the
focus
of
those
studies
was
not
to
apply
FA
as
a
biomarker.
In
total
we
found
53
studies,
covering
42
species
and
yielding
179
relationships
that
met
the
search
criteria
(Table
1).
In
order
to
include
studies
with
different
test
statistics,
we
used
correlation
coefficients
that
we
then
used
to
estimate
effect
sizes
(Cooper,
1998;
Hunter
and
Schmidt,
2004;
Rosenberg
et
al.,
2000).
Thus,
student
t-values
from
Levene’s
test,
F-values
from
ANOVAs,
chi-square
values,
Hedges’
effect
size
values
(g)
from
group
mean
comparisons
and
Spearman’s
correlation
coefficients
were
all
transformed
into
Pearson’s
correlation
coefficients
(r)
(Defife,
2009;
Rosenberg
et
al.,
2000).
If
direct
transformations
were
not
available,
we
transformed
the
p-values
into
Z-scores
and
then
transformed
the
values
into
our
common
summary
statistic
(r)
(Leung
and
Forbes,
1996;
Rosenberg
et
al.,
2000).
Author's personal copy
220
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
Table
1
Studies
on
the
effect
of
environmental
stress
on
fluctuating
asymmetry
included
in
the
meta-analysis:
insect
species,
stressor,
stressor
type,
stressor-fitness
relationship,
trait
measurement,
number
of
traits,
experimental
environment,
sample
size
and
reference.
Species
Stressor
Stressor
type
Stressor-fitness
relationship
Trait
measurement
Number
of
traits
Experimental
environment
Sample
size
Reference
Apis
mellifera
Pesticides
Anthropogenic
No
effect
Metric
3
Field
180
Abaga
et
al.
(2011)
Drosophila
melanogaster Endosulfane
exposure Anthropogenic Negative
effect Metric
Meristic
2
Lab
220,
221
Antipin
and
Imasheva
(2001)
Cyrtodiopsis
dalmanni Temperature Natural Negative
effect Composite 1 Lab 120 Bjorksten
et
al.
(2001)
Hydropsyche
morosa
Water
pollution
Anthropogenic
No
effect
Metric
5
Field
40
Bonada
and
Williams
(2002)
Hydropsyche
exocellata
Water
pollution
Anthropogenic
No
effect
Metric
3
Field
165
Bonada
et
al.
(2005)
Coenagrion
puella Parasitism Natural
No
effect
Metric
Meristic
2
Field
148
Bonn
et
al.
(1996)
Lygus
hesperus Nutrition Natural Negative
effect Composite 1 Lab
3293
Brent
(2010)
Coenagrion
puella
Atrazine
Carbaryl
Endosulfan
Anthropogenic
Negative
effect
Metric
1
Lab
300
Campero
et
al.
(2007)
Cephus
cinctus
Norton Wheat
host
variation Natural Negative
effect Metric 5 Lab 171 Cárcamo
et
al.
(2008)
Copera
annulata
Temperature
Avermectin
B
Imidacloprid
Anthropogenic
Negative
effect
Metric
5
Lab
35–39
Chang
et
al.
(2007a)
Ceriagrion
sp.
Insecticide
Anthropogenic
Negative
effect
Metric
3
Lab
22
Chang
et
al.
(2007b)
Copera
annulata Avermectin
B
Imidacloprid
Anthropogenic No
effect Metric 8 Lab 44–58
Chang
et
al.
(2009)
Musca
vetustissima Avermectin
B Anthropogenic Negative
effect Metric 1 Field 420 Clarke
and
Ridsdillsmith,
1990
Chrysopa
perla
Water
pollution
Anthropogenic
Negative
effect
Meristic
1
Field
161
Clarke
(1993)
Haematobia
irritans Diflubenzuron
Anthropogenic
Negative
effect
Metric
2
Lab
109
Da
Silva
et
al.
(2004)
Hexagenia
rigida PCB Anthropogenic No
effect Metric
3
Field
30
Dobrin
and
Corkum,
1999
Scathophaga
stercoraria Ivermectin
Anthropogenic
No
effect
Metric
7,
12
Lab
30,
60
Floate
and
Coghlin
(2010)
Musca
domestica
L.
Ivermectin
Anthropogenic
Negative
effect
Metric
1
Lab
178
Floate
and
Fox
(2000)
Pararge
aegeria
L. Larval
density Natural Negative
effect Composite
1
Lab
280
Gibbs
and
Breuker
(2006)
Brevicoryne
brassicae
Lead
accumulation
in
radish
and
cabbage
plant
Copper
accumulation
in
radish
and
cabbage
plant
Anthropogenic
Negative
effect
Negative
effect
Metric
4
Lab
120
Görür
(2006)
Brevicoryne
brassicae Zinc
accumulation
in
radish
and
cabbage
plant
Cadmium
accumulation
in
radish
and
cabbage
plant
Anthropogenic Negative
effect Metric
4
Lab
120
Görür
(2009)
Drosophila
melanogaster
Lead
and
benzene
exposure
Anthropogenic
Meristic
1
Lab
960
Graham
et
al.
(1993)
Chironomus
riparius
Metal-contaminated
river
Anthropogenic
No
effect
Meristic
1
(per
site)
Field
Lab
620
(field)
334
(lab)
Groenendijk
et
al.
(1998)
Xanthocnemis
zealandica
Carbaryl
exposure
Anthropogenic
No
effect
Composite
1
Lab
72
Hardersen
et
al.
(1999)
Xanthocnemis
zealandica Carbaryl
exposure Anthropogenic No
effect Metric 3 Lab
141
Hardersen
(2000)
Xanthocnemis
zealandica
Short
term
pesticide
exposure
Anthropogenic
Negative
effect
Composite
Metric
Meristic
3
Lab
40
Hardersen
and
Frampton
(1999)
Xanthocnemis
zealandica
Carbaryl
exposure
Anthropogenic
No
effect
Meristic
Composite
Metric
3
Lab
57
Hardersen
and
Wratten
(1998)
Helicoverpa
punctigera Pyrethroid
esfenvelerate
exposure
Cold
temperature
exposure
Anthropogenic
Natural
Negative
effect Composite 4 Lab
73
Hoffmann
et
al.
(2002)
Author's personal copy
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
221
Table
1
(Continued
)
Species
Stressor
Stressor
type
Stressor-fitness
relationship
Trait
measurement
Number
of
traits
Experimental
environment
Sample
size
Reference
Nemuora
trispinosa
Temperature
Natural
Negative
effect
Meristic
Composite
2
Field
172
Hogg
et
al.
(2001)
Scathophaga
stercoraria
Temperature
Density
Natural
Negative
effect
Composite
4
Lab
64
Hosken
et
al.
(2000)
Drosophila
pachea
Cactus
host
variation
Natural
Negative
effect
Metric
2
Lab
84,
86
Hurtado
et
al.
(1997)
Drosophila
melanogaster
Nutrition
Natural
Negative
effect
Metric
3
Lab
539
Imasheva
et
al.
(1999)
Drosophila
melanogaster
Drosophila
buzzatti
Temperature
Natural
Negative
effect
Metric
Meristic
3
(per
species)
Lab
300
Imasheva
et
al.
(1997)
Drosophila
subobscura
Lead
exposure
Anthropogenic
No
effect
Metric
1
Lab
32
Kurbalija
et
al.
(2010)
Pterostichus
melanarius
Fungicides
exposure
Insecticide
exposure
Anthropogenic
Negative
effect
Metric
Meristic
3
Field
587
Labrie
et
al.
(2003)
Culex
pipiens
Temperature
Natural
Negative
effect
Metric
3
Lab
109,
341,
376
Mpho
et
al.
(2002)
Culex
quinquefasciatus
Density
Natural
Negative
effect
Metric
4
Lab
35,
144
Mpho
et
al.
(2000)
Culex
quinquefasciatus
Temperature
Anthropogenic
Negative
effect
Composite
6
Lab
11,
12,
19,
22,
24
Mpho
et
al.
(2001)
Argia
tinctipennis
Selys
Riparian
vegetation
removal
Natural
Negative
effect
Metric
Meristic
10
Field
70
Pinto
et
al.
(2012)
Drosophila
melanogaster Arsenic
Anthropogenic
No
effect
Merisitic
2
Lab
1229
Polak
et
al.
(2002)
Drosophila
melanogaster
Temperature
Lead
exposure
Natural
Anthropogenic
Negative
effect
No
effect
Meristic
4
Lab
686
Polak
et
al.
(2004)
Formica
pratensis
Retz.
Metal-contaminated
sites
near
smelting
plant
Anthropogenic
No
effect
Metric
4
Field
350
Rabitsch
(1997)
Chironomus
riparius
Seasonal
and
temporal
variation
urban
and
industrial
wastewater
Anthropogenic
No
effect
Metric
Meristic
Composite
4
Field
44
Servia
et
al.
(2004)
Eulaema
nigrita
Lepeletier
Euglossa
pleosticta
Dressler
Seasonal
variation
Anthropic
influence
Natural
Anthropogenic
No
effect
Negative
effect
Metric
4
Field
60
Silva
et
al.
(2009)
Drosophila
buzzatti
Drosophila
koepferae
Catcus
host
variation
Natural
Negative
effect
Composite
4
Lab
53,
55
Soto
et
al.
(2008)
Drosophila
antonietae
Plant
Host
Natural
Negative
effect
Composite
2
Lab
325
Soto
et
al.
(2010)
Pararge
aegeria
L.
Nutrition
Natural
Negative
effect
Composite
1
Lab
581
Talloen
et
al.
(2004)
Pleioplectron
simplex
Parasitism
Natural
No
effect
Metric
3
Lab
65
Thomas
et
al.
(1998a)
Pleioplectron
simplex
Parasitism
Natural
No
effect
Metric
1
Lab
32
Thomas
et
al.
(1998b)
Drosophila
melanogaster
Temperature
Natural
Negative
effect
Metric
1
Lab
35
Trotta
et
al.
(2005)
Dacus
dorsalis
Methyl
eugenol
exposure
Anthropogenic
Negative
effect
Metric
1
Field
48
Tsubaki
(1998)
Drosophila
ananassae
Temperature
Natural
Negative
effect
Metric
Meristic
5
Lab
600
Vishalakshi
and
Singh
(2008)
Parellipsidion
pachycercum
Celatoblatta
quicquemaculata
Sigaus
obelisci
Paprides
dugdali
Phaulacridium
marginale
Concephalus
bilineatus
Parasitism
Natural
No
effect
Metric
1
(per
species)
Lab
34
Ward
et
al.
(1998)
Scathophaga
stercoraria
Avermectin
B
Anthropogenic
No
effect
Metric
1
Field
113
Webb
et
al.
(2007)
Author's personal copy
222
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
2.2.
Classification
of
studies
We
first
classified
studies
by
whether
they
were
conducted
in
the
field
or
in
laboratory
conditions
to
test
if
studies
under
con-
trolled
conditions
detected
larger
effects
on
FA.
We
also
coded
studies
by
whether
or
not
the
environmental
stressor
was
anthro-
pogenic
to
test
if
studies
that
assessed
novel
stressors
detected
larger
effects
on
FA
compared
to
studies
that
tested
more
naturally
occurring
stressors.
To
address
if
studies
that
applied
a
multidi-
mensional
measurement
or
a
combination
of
trait
measurements
detected
larger
effects
of
the
stressor
on
FA,
we
categorized
stud-
ies
by
the
kind
of
trait
measured
and
the
associated
measurement
method:
meristic
(counts),
metric
(linear)
or
composite
(shape
and
size).
Finally,
because
lack
of
a
significant
effect
on
FA
in
some
studies
may
be
due
to
the
environmental
stressor
not
being
highly
stressful
for
the
study
population,
we
categorized
studies
by
whether
or
not
they
found
a
negative
effect
on
fitness-related
traits
(i.e.
fecundity,
reduction
in
body
size).
2.3.
Meta-analysis
We
calculated
effect
sizes
for
each
relationship
by
transforming
individual
summary
measures
(r)
with
Fisher-z-transformation
for-
mula
(Zr):
Zr
=1
2ln 1
+
r
1
r
and
variance
(vz):
vz=1
n
3
where
n
is
the
sample
size,
using
MetaWin
2.0
software
(Rosenberg
et
al.,
2000).
We
then
used
an
unstructured
random-effects
model
to
calculate
a
mean
effect
size
(ES)
across
all
studies
(Rosenberg
et
al.,
2000).
A
random
effects
model
allowed
us
to
account
for
vari-
ation
in
effect
sizes
not
explained
by
sampling
error.
This
approach
is
recommended
for
meta-analyses
involving
data
from
ecological
studies
because
the
true
effect
size
is
expected
to
randomly
vary
due
to
ecological
factors
not
tested
(Gurevitch
et
al.,
2001).
We
cal-
culated
95%
confidence
intervals
based
on
the
parametric
variance
estimate
of
the
mean
effect
size
at
the
2.5
and
97.5
percentile
values.
Additionally,
we
calculated
bootstrap
confidence
intervals
based
on
nonparametric
variance
around
a
sample
of
individual
effect
size
estimates
using
a
resampling
test
with
9999
iterations
to
provide
a
more
conservative
estimate
of
the
mean
effect
size
(Adams
et
al.,
1997;
Rosenberg
et
al.,
2000).
Mean
effect
sizes
were
considered
to
be
significantly
different
from
zero
(p
<
0.05)
if
their
confidence
intervals
did
not
include
zero
(Hunter
and
Schmidt,
2004).
To
test
for
different
sources
of
variation
amongst
studies
in
the
sign
and
magnitude
of
effect
of
the
relationship
between
the
environmental
stressor
and
FA,
we
classified
studies
based
on
explanatory
vari-
ables
(see
Section
2.2)
and
examined
between-group
heterogeneity
(Qb)
(Gurevitch
et
al.,
2001).
We
assessed
the
total
heterogeneity
(QT)
against
a
chi-square
distribution
to
test
how
consistent
the
variance
was
across
all
effect
sizes
(Rosenberg
et
al.,
2000).
A
signif-
icant
QTindicates
that
variance
amongst
effect
sizes
is
too
different
for
meaningful
comparison
and
suggests
that
other
explanatory
variables
should
be
investigated
(Cooper,
1998;
Gurevitch
et
al.,
2001).
We
also
visually
inspected
the
data
with
a
funnel
plot
of
effect
size
estimates
against
sample
size
(Fig.
1).
Additionally,
to
evalu-
ate
the
robustness
of
our
findings
we
applied
a
fail-safe
calculation
based
on
Rosenberg’s
method
that
provides
a
weighted
fail-safe
estimate
(Rosenberg,
2005).
A
large
fail-safe
number
relative
to
the
number
of
observed
studies
indicates
that
the
observed
results,
even
with
publication
bias,
can
be
treated
as
a
reliable
estimate
of
Fig.
1.
Funnel
plot
of
Fisher-z-transformed
effect
sizes
in
relation
to
log
10-
transformed-sample
size
with
line
to
indicate
mean
effect
size.
the
true
effect
(Rosenberg
et
al.,
2000).
Additionally,
we
can
detect
the
presence
of
publication
bias
as
a
negative
covariance
between
effect
size
and
sample
size
(Begg,
1994).
Thus,
we
ran
a
regression
analysis
of
effect
size
on
sample
size.
3.
Results
3.1.
Mean
effect
size
and
tests
for
publication
bias
We
squared
the
value
of
our
estimated
mean
effect
size
to
judge
the
proportion
of
variance
explained
by
environmental
stress
in
the
meta-analysis
(R2).
Thus,
effect
sizes
that
are
equal
to
or
less
than
0.10
are
considered
small
(corresponding
to
1%
of
the
variance
being
explained),
0.30
are
considered
medium,
and
0.50
indicates
a
large
effect
(Cohen,
1988).
Typically,
in
meta-analyses
of
ecolog-
ical
and
evolutionary
biology
studies,
the
proportion
of
variance
explained
by
the
main
factor
in
question
ranges
between
5
and
10%
(Møller
and
Jennions,
2002).
Overall,
the
mean
effect
size
across
all
studies
was
0.60
(95%
CI
=
0.50–0.70;
Bootstrap
CI
=
0.52–0.68),
indicating
that
environ-
mental
stressors
accounted
for
36%
of
the
variance
in
FA.
Variance
across
individual
effect
sizes
was
non-significant
(QT=
116.6;
d.f.
=
178;
2=
1.00).
Thus,
with
the
overall
large
effect,
we
can
expect
FA
to
be
highly
affected
by
environmental
stressors
and
therefore
serve
as
a
reliable
biomarker
of
environmentally
induced
developmental
instability.
Our
weighted
fail-safe
estimate
was
76,
suggesting
our
find-
ings
are
unlikely
to
be
eliminated
by
publication
bias.
Furthermore,
our
regression
analysis
found
effect
size
increased
with
sample
size
(F
=
4.84,
d.f.
=
1,
177,
r2=
0.03,
p
=
0.03,
slope
(SE)
=
1.89
×
101
(8.63
×
102)),
which
is
contrary
to
the
expectations
of
publication
bias.
3.2.
Classification
of
studies
We
detected
a
significant
difference
between
estimates
that
demonstrated
a
negative
effect
of
the
stressor
on
fitness-related
traits
compared
to
those
that
showed
no
effect
of
the
stressor
(Qb=
9.56,
d.f.
=
1,
p
<
0.01).
Studies
where
the
stressor
had
a
signif-
icant
effect
on
fitness-related
traits
also
tended
to
have
the
largest
effect
sizes
(effect
size
(ES)
=
0.70,
d.f.
=
123,
95%
CI
=
0.58–0.82,
Bootstrap
CI
=
0.60–0.81,
n
=
34,
k
=
123)
when
compared
to
stud-
ies
that
did
not
find
such
an
effect
(ES
=
0.37,
d.f.
=
55,
95%
CI
=
0.19–0.55,
Bootstrap
CI
=
0.29–0.45,
n
=
18,
k
=
56;
Fig.
2).
Thus,
our
estimation
of
the
sensitivity
of
FA
as
a
biomarker
of
exposure
to
an
environmental
stress
or
is
likely
to
be
conservative,
as
studies
that
did
not
find
an
effect
of
the
stressor
on
FA
might
have
dealt
with
a
factor
that
was
not
affecting
the
study
species
at
all.
When
we
grouped
studies
by
whether
they
were
conducted
under
field
or
laboratory
conditions,
we
detected
significant
Author's personal copy
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
223
Fig.
2.
Forest
plot
of
mean
effect
sizes
categorized
by
the
effect
of
stressor
on
fitness-related
traits
(negative
effect
vs.
no
effect)
in
order
of
increasing
effect
size.
Effect
sizes
are
z-transformed
Pearson
correlation
coefficient
estimates
and
shown
with
95%
confidence
intervals.
QT=
120,
d.f.
=
178,
2=
1.00.
Fig.
3.
Forest
plot
of
mean
effect
sizes
categorized
by
experimental
environment
(field
vs.
lab)
in
order
of
increasing
effect
size.
Effect
sizes
are
z-transformed
Pearson
correlation
coefficient
estimates
and
shown
with
95%
confidence
intervals.
QT=
118.7,
d.f.
=
178,
2=
1.00.
Table
2
Mean
effect
sizes
(ES),
confidence
intervals
(CI),
number
of
studies
(n),
number
of
relationships
presented
in
the
study
(k),
heterogeneity
among
studies
(QT),
degrees
of
freedom
(d.f.
=
k
1),
chi-square
distribution
(2)
grouped
by
environmental
stressor
type:
anthropogenic
and
natural.
Note
that
some
studies
contributed
data
for
more
than
one
category
so
n-values
do
not
sum
up
to
total
number
of
studies
in
meta-analysis.
Stressor
type
Mean
ES
95%
CI
Bootstrap
CI
n
k
QTd.f.
2
All
studies
0.60
0.50,
0.70
0.52,
0.68
54
179
121.2
178
1.00
Anthropogenic
0.56
0.43,
0.70
0.48,
0.66
30
102
Natural
0.64
0.49,
0.79
0.50,
0.79
22
77
variation
across
(Qb=
4.91,
d.f.
=
1,
p
=
0.01),
with
laboratory-
based
studies
having
larger
effect
sizes
(ES
=
0.66,
d.f.
=
131,
95%
CI
=
0.55–0.78,
Bootstrap
CI
=
0.56–0.77,
n
=
39,
k
=
132)
compared
to
field
studies
(ES
=
0.41,
d.f.
=
46,
95%
CI
=
0.22–0.61,
Bootstrap
CI
=
0.32–0.50,
n
=
14,
k
=
47)
suggesting
that
the
conditions
under
which
the
study
was
conducted
explain
some
of
the
variation
amongst
studies
(Fig.
3).
Next,
we
tested
if
the
type
of
envi-
ronmental
stressor
accounted
for
the
variation
in
FA
response
and
did
not
detect
any
significant
variation
according
to
whether
the
stressor
was
naturally
occurring
or
anthropogenic
(Qb=
0.60,
d.f.
=
1,
p
=
0.34;
Table
2).
Finally,
we
did
not
find
a
significant
dif-
ference
between
studies
classified
by
the
kind
of
trait
that
they
measured
(Qb=
0.72,
d.f.
=
2,
p
=
0.58;
Table
3).
4.
Discussion
Our
study
used
a
meta-analytical
approach
to
assess
the
reliabil-
ity
of
FA
as
a
biomarker
of
environmentally
induced
developmental
instability,
and
to
identify
features
of
different
studies
that
might
explain
variation
in
FA
response
to
changes
in
environmental
Table
3
Mean
effect
sizes
(ES),
confidence
intervals
(CI),
number
of
studies
(n),
number
of
relationships
presented
in
the
studies
(k),
heterogeneity
among
studies
(QT),
degrees
of
freedom
(d.f.
=
k
1),
chi-square
distribution
(2)
grouped
by
measures
of
traits:
metric,
meristic
and
composite.
Note
that
some
studies
contributed
data
for
more
than
one
category
so
n-values
do
not
sum
up
to
total
number
of
studies
in
meta-analysis.
Trait
measure
Mean
ES
95%
CI
Bootstrap
CI
n
k
QTd.f.
2
All
studies
0.60
0.50,
0.70
0.52,
0.68
54
179
116.6
178
1.00
Metric
0.57
0.44,
0.69
0.47,
0.67
19
111
Meristic 0.68
0.43,
0.93
0.44,
0.94
5
30
Composite
0.63
0.40,
0.85
0.51,
0.75
6
38
Author's personal copy
224
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
quality.
Overall,
we
found
that
FA
can
serve
as
a
sensitive
biomarker,
with
environmental
stress
having
a
very
large
effect
on
FA
response
in
our
sample
of
studies.
Particularly,
studies
that
verified
a
negative
effect
of
the
stressor
on
fitness-related
traits
detected
significantly
larger
effects.
Studies
conducted
under
the
controlled
conditions
of
a
lab
detected
comparatively
larger
effects
of
the
environmen-
tal
stressor
on
FA
than
studies
of
wild
populations.
Other
study
factors
such
as
method
of
measuring
traits
and
the
nature
of
the
stressor
did
not
significantly
explain
variation
amongst
studies,
suggesting
recent
improvement
in
analytical
tools
and
sampling
considerations
have
standardized
and
enhanced
the
rigor
of
FA
analysis
across
insect
studies
(Palmer
and
Strobeck,
2003).
As
a
biomarker,
FA
is
particularly
advantageous
due
to
the
ease
in
identifying
optimal
levels
(i.e.
perfect
symmetry)
compared
to
physiological
biomarkers.
Determining
the
optimum
level
of
physiological
biomarkers
is
complicated
for
two
reasons:
one,
the
up-regulation
of
these
biomarkers
usually
occurs
at
the
expense
of
their
availability
for
other
functions.
For
example,
antioxidants
in
birds
play
an
important
role
in
both
immune
function
and
the
col-
oration
of
secondary
sexual
characters
and
have
recently
been
used
as
an
indicator
of
exposure
to
radiological
contaminants
(Bonisoli
Alquati
et
al.,
2010;
Peters,
2007).
Studies
have
demonstrated
how
reduced
availability
of
carotenoids
due
to
increased
immune
activ-
ity
resulted
in
paler
color
expression
in
secondary
sexual
characters
(Aguilera
and
Amat,
2007;
Alonso-Alvarez
et
al.,
2004).
Thus,
with
multiple
biological
functions
influencing
the
levels
of
antioxidants,
it
may
be
difficult
to
determine
when
levels
indicate
a
beneficial
or
detrimental
response
to
a
stressor.
Two,
while
short-term
response
of
some
physiological
biomarkers
to
a
stressor
is
adaptive,
it
may
prove
to
be
detrimental
in
the
long
term
if
the
stressor
persists
(Dhabhar,
1999).
For
instance,
inflammatory
immune
responses
are
known
to
cause
tissue
damage
over
the
long
term
and
pro-
longed
exposure
to
glucocorticoids
causes
significant
damage
to
the
biological
system
(Maccari
et
al.,
2003).
Additionally,
FA
is
a
less
invasive
measure
of
quality
com-
pared
to
more
invasive
techniques
that
require
removing
tissue
or
killing
the
animal.
For
example,
one
of
the
better-known
molec-
ular
biomarkers,
cytochrome
P450,
has
proven
to
be
a
very
reliable
indicator
of
exposure
to
organochlorine
contaminants
(Sarkar
et
al.,
2006).
However,
the
use
of
this
biomarker
has
traditionally
required
the
harvesting
of
liver
tissue
from
dead
animals,
presenting
legal
and
ethical
concerns
for
conservation
studies
(Miller,
2003).
In
the
case
of
heat
shock
proteins,
technical
issues
such
as
the
availabil-
ity
of
antibodies
for
specific
species
limits
the
use
of
these
stress
proteins
for
environmental
assessment
(Sørensen,
2010).
The
fitness
consequences
associated
with
elevated
levels
of
FA
in
addition
to
its
sensitivity
to
environmental
stress
is
important
for
conservation
studies
(Depledge
and
Fossi,
1994;
Møller,
1997).
Schmeller
et
al.
(2011)
found
that
variance
in
population-wide
FA
in
the
wing
veins
of
the
Mountain
Apollo
Butterfly
(Parnassius
apollo),
an
endangered
butterfly
species,
was
comparable
to
FA
variance
of
pre-threatened
populations
13
years
after
management
prac-
tices
went
into
effect,
supporting
the
use
of
FA
as
an
assessment
tool,
although
the
potential
effects
on
FA
related
to
inbreeding
in
small
populations
must
also
be
assessed.
Additionally,
Monna
et
al.
(2011)
used
FA
to
assess
developmental
quality
in
wild
brown
trout
(Salmo
trutta
fario)
exposed
to
heavy
metal
pollution
in
a
protected
area
that
historically
was
a
site
for
heavy
mining
activity.
The
high
levels
of
FA
in
populations
in
relation
to
high
metal
contamination
emphasized
the
importance
of
considering
an
area’s
contamination
history
when
implementing
long-term
management
plans.
The
association
of
the
increased
FA
with
negative
effects
on
fitness-related
traits
emphasizes
the
importance
of
considering
the
evolutionary
potential
of
populations
and
may
offer
insight
into
why
results
in
the
application
of
FA
have
been
inconsistent
in
the
literature.
There
is
increasing
recognition
for
the
incorporation
of
evolutionary
ecology
in
conservation
management
due
to
the
need
for
understanding
how
global
environmental
changes
and
manage-
ment
practices
influence
a
species’
adaptive
response
(Davis
et
al.,
2005;
Stockwell
et
al.,
2003).
Hoffmann
and
Willi
(2008)
presented
evidence
demonstrating
changes
in
candidate
loci
responsible
for
adaptive
responses,
such
as
the
alcohol
dehydrogenase
gene
(Adh)
in
the
fruit
fly
(Drosophila
melanogaster)
and
the
glycerate
dehy-
drogenase
locus
(Gly)
in
Pi˜
non
Pine
(Pinus
edulis),
with
various
environmental
stressors
and
climate
change.
Thus,
interpretation
of
FA
should
be
considered
within
the
limits
of
the
population’s
adaptive
potential
and
current
understanding
of
a
population’s
life
history.
For
example,
Hogg
et
al.
(2001)
looked
at
FA
as
a
biomarker
of
water
temperature
stress
while
considering
the
underlying
pop-
ulation
genetics
of
the
stonefly
(Nemoura
trispinosa)
and
detected
a
negative
correlation
between
levels
of
FA
and
heterozygosity.
Lens
et
al.
(2000)
found
similar
results
when
they
assessed
FA
as
a
biomarker
of
habitat
disturbance
in
populations
of
Taita
thrush
(Turdus
helleri)
with
reduced
heterozygosity.
Levels
of
FA
were
more
pronounced
in
highly
disturbed
areas
but
weak
under
less
disturbed
conditions.
We
predicted
that
lab
studies
would
detect
larger
effects
of
the
environmental
stressor
on
FA
as
a
result
of
fewer
factors
interfering
with
the
strength
of
the
relationship.
A
previous
meta-analysis
on
FA
and
environmental
stress
suggested
that
the
lack
of
biological
validation
of
the
stressor
could
dilute
the
strength
of
the
association
(Leung
and
Forbes,
1996).
This
is
more
difficult
to
determine
in
wild
populations
that
are
exposed
not
only
to
the
stressor
in
question
but
to
additional
factors
that
have
unknown
effects
on
the
study
organism.
Additionally,
our
finding
that
lab-based
studies
detected
larger
magnitudes
of
effect
of
the
environmental
factor
on
FA,
fur-
ther
emphasizes
the
importance
of
considering
the
population’s
adaptive
potential
in
relation
to
its
response
to
stress.
For
instance,
stronger
effects
of
stressors
on
FA
may
also
be
attributed
to
reduced
genetic
diversity
for
stress
tolerance
in
lab
populations
while
wild
populations
with
more
genetic
diversity
have
more
adaptive
poten-
tial
and
comparatively
lower
FA
as
a
result.
Frankham
(2005)
found
that
loss
of
genetic
variation
in
D.
melanogaster
raised
in
controlled
conditions
reduced
stress
resistance
and
significantly
suppressed
the
evolutionary
adaptive
potential
in
stressful
environments
when
populations
were
re-introduced
into
the
field.
Thus,
lab
studies
such
as
artificial
selection
experiments
may
provide
insight
into
understanding
FA
response
to
environmental
stressors
in
relation
to
adaptation
and
trade-offs
with
stress
resistance
and
fitness-
related
traits
in
vulnerable
populations
(Bijlsma
and
Loeschcke,
2005).
Our
study
showed
larger
strengths
of
association
between
environmental
stress
and
FA
compared
to
previously
published
meta-analyses,
which
included
multiple
taxa
(Leung
and
Forbes,
1996;
Hogg
et
al.,
2001).
One
possible
reason
may
be
that
primary
studies
on
mammals
or
fish,
where
larger
sample
sizes
are
difficult
to
obtain,
are
more
likely
to
be
hampered
by
measurement
error,
other
asymmetry
types
and
trait
choice,
compared
to
studies
on
insects
(Allenbach,
2010).
Another
consideration
is
that
the
life
his-
tory
of
many
insects
leaves
them
comparatively
more
susceptible
to
environmental
changes,
making
the
taxa
a
more
suitable
model
for
using
FA
to
evaluate
environmental
quality.
Additional
studies
on
how
FA
responds
within
taxa
will
help
better
understand
the
general
use
and
interpretation
of
FA
across
taxa.
Finally,
our
study
did
not
find
a
significant
difference
between
naturally
occurring
and
anthropogenic
stressors
in
causing
FA.
This
is
in
contrast
to
a
previous
study
that
detected
a
significant
associ-
ation
between
toxic
stressors
and
FA
(Hogg
et
al.,
2001).
Our
results
suggest
that
the
duration
of
exposure
may
have
greater
influence
on
FA.
However,
formulating
predictions
are
difficult
given
that
prolonged
exposure
to
a
stressor
can
cause
either
developmen-
tal
selection
for
symmetry
that
would
reduce
levels
of
FA
in
the
Author's personal copy
D.E.
Beasley
et
al.
/
Ecological
Indicators
30
(2013)
218–226
225
population
or
increase
the
population’s
mutational
load
that
would
impair
development
and
thus
increase
FA.
Studies
that
look
at
the
change
in
level
of
FA
as
the
selection
pressure
increases
could
help
determine
the
role
of
prolonged
exposure
when
using
FA
to
assess
environmental
stress
(Polak
et
al.,
2002).
In
conclusion,
the
use
of
FA
as
a
biomarker
may
be
most
informa-
tive
when
a
negative
effect
of
the
stressor
exists
on
fitness-related
traits,
verifying
that
the
environmental
factor
in
question
is
indeed
a
stressor.
To
further
our
understanding
on
the
use
and
limitations
of
FA
as
a
biomarker,
future
studies
may
wish
to
clarify
how
FA’s
response
to
environmental
stress
varies
over
the
course
of
expo-
sure
within
a
population.
Such
a
study
could
potentially
expand
into
multi-generational
responses
of
FA
to
a
stressor,
and
insect
models
would
be
ideal
(e.g.
Beasley
et
al.,
2012).
This
is
particu-
larly
important
for
contaminants
such
as
oil
spills,
radiological
and
metal
pollutants
that
are
expected
to
persist
in
the
environment
for
long
periods
of
time.
Studies
looking
at
FA
in
relation
to
adaptive
responses
of
a
population
will
provide
additional
insight
to
how
FA
co-varies
with
stress
in
the
context
of
evolutionary
constraints
and
plasticity.
Acknowledgements
The
authors
would
like
to
thank
Anders
P.
Møller,
Shane
Welch
and
Jayme
Waldron
for
their
helpful
comments
that
improved
the
manuscript.
Two
reviewers
provided
constructive
criticism
on
a
previous
version
of
this
paper.
This
study
was
partially
funded
through
a
GAANN
Fellowship
to
DEB.
References
Abaga,
N.O.Z.,
Alibert,
P.,
Dousset,
S.,
Savadogo,
P.W.,
Savadogo,
M.,
Sedogo,
M.,
2011.
Insecticide
residues
in
cotton
soils
of
Burkina
Faso
and
effects
of
insecticides
on
fluctuating
asymmetry
in
honey
bees
(Apis
mellifera
Linnaeus).
Chemosphere
83,
585–592.
Adams,
D.C.,
Gurevitch,
J.,
Rosenberg,
M.S.,
1997.
Resampling
tests
for
meta-analysis
of
ecological
data.
Ecology
78,
1277–1283.
Adams,
S.M.,
Giesy,
J.P.,
Tremblay,
L.A.,
Eason,
C.T.,
2001.
The
use
of
biomarkers
in
ecological
risk
assessment:
recommendations
from
the
Christchurch
conference
on
biomarkers
in
ecotoxicology.
Biomarkers
6,
1–6.
Aguilera,
E.,
Amat,
J.A.,
2007.
Carotenoids,
immune
response
and
the
expression
of
sexual
ornaments
in
male
greenfinches
(Carduelis
chloris).
Naturwissenschaften
94,
895–902.
Allenbach,
D.M.,
2010.
Fluctuating
asymmetry
and
exogenous
stress
in
fishes:
a
review.
Rev.
Fish
Biol.
Fish.
21,
355–376.
Alonso-Alvarez,
C.,
Bertrand,
S.,
Devevey,
G.,
Gaillard,
M.,
Prost,
J.,
Faivre,
B.,
Sorci,
G.,
2004.
An
experimental
test
of
the
dose-dependent
effect
of
carotenoids
and
immune
activation
on
sexual
signals
and
antioxidant
activity.
Am.
Nat.
164,
651–659.
Antipin,
M.I.,
Imasheva,
A.G.,
2001.
Genetic
variability
and
fluctuating
asymmetry
of
morphological
traits
in
Drosophila
melanogaster
reared
on
a
pesticide-containing
medium.
Genetika
37,
247–252.
Arnqvist,
G.,
Wooster,
D.,
1995.
Meta-analysis
synthesizing
research
findings
in
ecology
and
evolution.
Trends
Ecol.
Evol.
10,
236–240.
Beasley,
D.E.,
Bonisoli
Alquati,
A.,
Welch,
S.M.,
Møller,
A.P.,
Mousseau,
T.A.,
2012.
Effects
of
parental
radiation
exposure
on
developmental
instability
in
grasshop-
pers.
J.
Evol.
Biol.
25,
1149–1162.
Begg,
C.B.,
1994.
Publication
bias.
In:
Cooper,
H,
Hedges,
L.V.
(Eds.),
The
Handbook
of
Research
Synthesis.
Russel
Sage
Foundation,
New
York,
pp.
399–409.
Bijlsma,
R.,
Loeschcke,
V.,
2005.
Environmental
stress,
adaptation
and
evolution:
an
overview.
J.
Evol.
Biol.
18,
744–749.
Bjorksten,
T.A.,
Pomiankowski,
A.,
Fowler,
K.,
2001.
Temperature
shock
during
devel-
opment
fails
to
increase
the
fluctuating
asymmetry
of
a
sexual
trait
in
stalk-eyed
flies.
Proc.
R.
Soc.
B
268,
1503–1510.
Bonada,
N.,
Williams,
D.D.,
2002.
Exploration
of
the
utility
of
fluctuating
asymme-
try
as
an
indicator
of
river
condition
using
larvae
of
the
caddisfly
Hydropsyche
morosa
(Trichoptera:
Hydropsychidae).
Hydrobiologia
481,
147–156.
Bonada,
N.,
Vives,
S.,