Ecology, 91(9), 2010, pp.

2628–2637
Ó 2010 by the Ecological Society of America

Induced responses to competition and herbivory:

natural selection on multi-trait phenotypic plasticity
KARINA BOEGE1
Instituto de Ecologı´a, Universidad Nacional Autónoma de Me´xico (UNAM), Apartado Postal 70-275,
Ciudad Universitaria, Mexico D.F. 04510 Mexico

Abstract. Herbivory and competition are two of the most common biotic stressors for
plants. When occurring simultaneously, responses to one interaction can constrain the
induction of responses to the other interaction due to resource limitation and other interactive
effects. Thus, to maximize fitness when interacting with competitors and herbivores, plants are
likely to express particular combinations of plastic responses. This study reports the
interactive effects of herbivory and competition on responses induced in Tithonia tubaeformis
plants and describes how natural selection acts on particular plastic responses and on their
different combinations. Competition induced a stem elongation response, expressed through
an increase in height and mean internode length, together with a decrease in basal diameter.
Interestingly, realized resistance increased in both competition and herbivory treatments,
suggesting a plastic response in both constitutive and induced resistance traits. Particular
combinations of plastic responses defined three plant phenotypes: vigorous, elongated, and
resistant plants. The ecological context in which plants grew modified the traits and the
particular combinations of plastic responses that were favored by selection. Vigorous plants
were favored by selection in all environments, except when they were damaged by herbivores
in the absence of neighbors. The combination of responses defining an elongated plant
phenotype was favored by selection in crowded conditions. Resistance was negatively selected
in the absence of competition and herbivory but favored in the presence of both interactions.
In addition, contextual analyses detected that population structure in heterogeneous
environments can also influence the outcomes of selection. These findings suggest that
natural selection can act on particular combinations of plastic responses, which may allow
plants to adjust their phenotypes to those that promote greater fitness under particular
ecological conditions.
Key words: competition; herbivory; multi-trait phenotypic plasticity; natural selection; Pedregal San
Ángel Reserve, Mexico City, Mexico; resistance; sunflower; Tithonia tubaeformis.

INTRODUCTION simultaneously to different types of interactions (e.g.,

Over their lifetimes, plants interact simultaneously
with a great number of species, each of which can have
positive or negative effects on their fitness. Because the
expression of traits involved in the attraction of
mutualist species or the deterrence of antagonist species
can be costly, and because the occurrence of each
interaction is somewhat unpredictable (Karban et al.
1999), plants often minimize costs by inducing such
traits when reliable cues are perceived. Although the
adaptive value and natural selection of many of these
plastic responses has been described (Schmitt et al. 1995,
Agrawal 1998, Baldwin 1998), most studies have
assessed such fitness consequences when plants face
only one type of interaction (e.g., herbivory), simplifying
the ecological context in which plant responses are
selected. However, when plants need to respond
Manuscript received 26 March 2009; revised 14 January
2010; accepted 22 January 2010. Corresponding Editor: P. M.
Kotanen.
1 E-mail: kboege@servidor.unam.mx
2628
mutualistic, competitive, etc.), resource allocation trade-
offs, pleiotropic effects, and other genetic, biochemical,
and/or ecological constraints can either promote syner-
gism or establish conflicts among responses to different
species, conditioning their costs and adaptive value
(Strauss et al. 1999, Cipollini 2002, Agrawal 2004, Adler
et al. 2006). Thus, to understand the evolution of plant
responses to the environment within a community
context, a diffuse selection approach is needed (Stinch-
combe and Rausher 2001, Tiffin 2002, Strauss and Irwin
2004).
Herbivory and competition are two of the most
common biotic stressors that plants face simultaneously
(Guerevich et al. 2000, Hambäck and Beckerman 2003,
Agrawal 2004, Cipollini 2004) and whose independent
negative impacts on plant fitness have been well
described (Crawley 1997). To reduce such impacts,
many plants are able to express several plastic responses.
For example, in the presence of herbivores, plants have
the ability to induce resistance traits that decrease the
amount of herbivore damage (Karban and Baldwin
September 2010 MULTI-TRAIT PHENOTYPIC PLASTICITY
1997). Plants may also express tolerance once the
damage has occurred, which allows them to reduce its
negative impacts (Stowe et al. 2000). Moreover, when
plants are nutrient limited by competitors and they are
unable to grow, they often can increase their constitutive
levels of resistance by accumulating carbon-based
secondary compounds (Mihaliak and Lincoln 1989,
Stamp 2003, Boege and Dirzo 2004). In the case of
competition, plants can express traits that minimize the
intensity and/or negative impacts of this antagonistic
interaction. For example, under light competition, many
plants activate a stem-elongation response (Dudley and
Schmitt 1996) and reduce the number of leaves (Dorn et
al. 2000). Similarly, many plants reallocate resources to
root biomass under nutrient competition (Callaway et
al. 2003), and as a consequence, change their morphol-
ogy (Agrawal and Van Zandt 2003). Most of these
responses, however, have been described independently
for environments with herbivory or competition (Kar-
ban et al. 1989, Cipollini and Bergelson 2001, Siemens et
al. 2002, 2003, Callaway et al. 2003, Kurashige and
Agrawal 2005, McGuire and Agrawal 2005). If limiting
resources impose a trade-off between herbivore- and
competition-induced responses, then a plant’s simulta-
neous interaction with herbivores and competitors could
constrain the adaptive evolution of phenotypic plastic
responses (Cipollini 2002, 2004, Callaway et al. 2003,
Siemens et al. 2003, Agrawal et al. 2006), promoting
diffuse selection on such traits.
Based on the idea that defense is costly and that there
should be resource allocation trade-offs between growth
and defense (Herms and Mattson 1992, Cipollini 2002),
the Active Response Model and the Compensatory
Continuum Hypothesis predict, respectively, that plants
growing at low densities should have more resources to
either induce resistance traits (Karban et al. 1989) or
tolerate herbivore damage (Maschinski and Whitham
1989) than plants in crowded conditions. Evidence from
many studies supports these two hypotheses (Bergelson
1994, Baldwin and Hamilton 2000, Weis and Hochberg
2000, Cipollini and Bergelson 2001, Tiffin 2002,
Kurashige and Agrawal 2005), suggesting that defenses
become more costly under stressful or crowded envi-
ronments. Furthermore, a recent study has demonstrat-
ed that even in the absence of resource limitation, the
perception of the proximity of neighbors is enough to
down-regulate gene expression of several induced
chemical defenses (Izaguirre et al. 2006). However,
recent evidence has also shown a decrease in costs of
resistance and tolerance in crowded conditions, relative
to environments without competition (Siemens et al.
2002, 2003, Cipollini 2007, 2010). From these studies,
the Defense Stress Benefit Hypothesis has been pro-
posed, which states that additional functions of defen-
sive traits may benefit plants under competition (Inderjit
and Del Moral 1997, Lankau and Strauss 2007),
reducing the total costs of such traits and promoting
their expression under crowded conditions (Siemens et
2629
al. 2003, Agrawal 2004, Cipollini 2007). Whether stress
decreases or increases costs of resistance is likely to
depend on the nature of the limiting resources for plant
growth and the impacts of herbivory on a plant’s ability
to acquire resources (Wise and Abrahamson 2005).
Given the interactive effects of competition and
herbivory on phenotypic plastic responses, and because
more than one response is usually promoted by each
interaction, it is likely that, to maximize fitness, plants
actually express particular combinations of plastic traits,
depending on the ecological context in which they grow.
If the simultaneous expression of such set of plastic
traits is indeed adaptive, we could expect to observe
natural selection to favor a particular combination of
such traits, instead of the independent expression of
each response. This correlative selection could, however,
be constrained by the correlation structure among traits
(Ordano et al. 2008). Although some studies have
assessed the selective impact of the simultaneous
occurrence of competition and herbivory on phenotypic
plastic traits (Tiffin 2002, Agrawal and Van Zandt
2003), there are currently no studies assessing natural
selection on the combined expression of phenotypic
plastic traits induced simultaneously by both competi-
tion and herbivory. Within this context, this paper
describes how natural selection acts on particular
combinations of plastic phenotypes, depending on the
ecological context in which plants grow. In particular,
the aim of this work was to (1) evaluate the simulta-
neous expression of plant responses induced by intra-
specific competition and herbivory, (2) determine
whether competition constrains or enhances the ability
of plants to respond to herbivores, and (3) evaluate
whether natural selection acts on each independent
response and/or on particular combinations of correlat-
ed induced responses, depending on the ecological
context in which they are expressed.
METHODS
Study site
This study was conducted at the Pedregal San Ángel
Reserve (PSAR), located within Mexico City (19820 0 3300
N, 99808 0 2600 W), at 2300 m above sea level. The
predominant vegetation of this area is xerophylous
shrubland. Annual precipitation ranges from 700 to 900
mm and is concentrated between May and October
(Rzedowski 1954).
Study system
Tithonia tubaeformis (Asteraceae) is a wild sunflower
distributed in the American continent across mountains,
deserts, and tropical dry forests, and it is common along
roadsides, grasslands, and open areas (McVaugh 1984).
Plants of T. tubaeformis grow in dense patches and are
eaten by a wide array of generalist herbivores (orthop-
terans, coleopterans, and lepidopterans) and by a
specialist Chrysomelid beetle (Zygograma signatipennis),
which can cause up to 90% of leaf area damage (K.
2630 KARINA BOEGE
Boege, personal observation). A T. tubaeformis popula-
tion was located within the PSAR, with plants growing
at high densities (;180 seedlings/m2) in discrete patches
separated by 2–20 m.
On June 2006, seedlings had developed their first pair
of true leaves. Within each of nine patches, 32 small
plots of 30 3 30 cm were randomly selected, in which
plants were tagged and subjected to either or both
competition and herbivory treatments with a 2 3 2 fully
factorial design. Each combination of both treatments
had 15 plants within each patch (N ¼ 60 plants/patch).
Plants with no competition were left without neighbors
(1 plant/plot, with a total of 30 noncompetition plots/
patch) by weeding all plants within a radius of 0.5 m
around the focal plant. During the rest of the growing
season, weeding was continued to maintain a noncom-
petitive environment. Plants under competition were
concentrated in two high-density plots (15 plants/plot,
with a total of 2 competition plots/patch). Plants with
early-season herbivory were left unmanipulated, where-
as plants without early-season herbivory were sprayed
with a contact insecticide (Sevin, Bayer, Mexico City,
Mexico) during 10 days. Enough insecticide was applied
to wet all leaves, but dripping was avoided (so watering
control plants was not necessary). Due to time
constraints, levels of herbivory after the induction
period were not assessed, but on average, herbivores
damaged 54% 6 0.01% (mean 6 SE) of the leaves of
plants under this treatment.
Resistance to herbivory was indirectly estimated using
the performance of the generalist herbivore Spodoptera
exigua. This indirect measure was influenced by all
constitutive and induced traits affecting herbivore
feeding abilities, thus from hereafter it will be referred
to as ‘‘realized resistance.’’ Because no previous studies
are available on T. tubaeformis resistance traits, there is
no information regarding the biochemical and physical
factors influencing estimated realized resistance. We
know however, that water, nitrogen, and phosphorus
content have no influence on herbivore performance (K.
Boege and A. Hernandez, unpublished data). Moreover,
sesquiterpene lactones have been found to be the main
constituents of glandular trichomes with an antifeedant
activity on caterpillars in the related species, T.
diversifolia (Ambrosio et al. 2008). Current research is
assessing whether these compounds are also responsible
for reducing herbivore performance in T. tubaeformis.
Any difference in realized resistance between plants with
and without early-season herbivory (but no competi-
tion) was interpreted as a result of the induction of
resistance traits after herbivore damage (Kurashige and
Agrawal 2005). In contrast, differences in realized
resistance between plants with and without competition
(but no herbivory) were interpreted as a result of
changes in constitutive resistance traits associated with
resource limitation mediated by competition (Cipollini
and Bergelson 2001). Finally, the contrast between
realized resistance of control plants and individuals
Ecology, Vol. 91, No. 9
interacting with both herbivores and competitors was
considered to be the result of an increase in both induced
and constitutive resistance traits. Third-instar larvae
were weighed and placed on Petri dishes, each contain-
ing one new leaf (produced after insecticide application
was stopped) of a plant assigned to one of the
treatments. A wet cotton ball was placed on the petiole
of leaves to preserve moisture during the next days.
After four days, larvae were weighed and their
performance was estimated as (final mass  initial
mass)/initial mass. Realized resistance of plants was
estimated as 1/performance. This magnitude represented
all those traits that reduced herbivore performance,
where high values indicate proportionally greater
resistance than low values.
Initial measurements of basal diameter, height, and
number of leaves were recorded for each plant. After 15
days, basal diameter, stem elongation, internode length,
and number of new leaves were quantified to assess
growth-related responses to competition and herbivory
treatments, as the difference of final vs. initial measure-
ments. In the case of internode length, which had no
initial measurement, the plastic response was inferred by
comparing plants within each treatment with the control
group. When plants reached maturity two months later,
their reproductive effort was estimated by counting the
number of flower heads. Five randomly selected flower
heads per plant were collected and their diameters were
measured to have an estimate of mean biomass
allocation to reproductive structures. Under the as-
sumption that the size of flower heads and the number of
seeds per flower head are positively correlated, a
component of fitness (henceforth ‘‘fitness’’) was estimat-
ed as the product of the number of flower heads 3 mean
flower head basal area.
Statistical analyses
In contrast with plants without competition within
each patch, plants within the competition plots were not
independent. Therefore, the average value of plant
responses in each combination of both treatments was
calculated within each patch. These mean values were
used as replicates (N ¼ 9 per treatment) in a MANOVA
and subsequent univariate ANOVAs to assess whether
treatments promoted different plant responses. To meet
assumptions of normality and homogeneity of variances,
these values were ln-transformed (new leaves, internode
length, and realized resistance) or arcsine square-root
(stem elongation) transformed. In these models, compe-
tition, herbivory, and their interaction were considered
as class variables. Post hoc comparisons were performed
to assess differences among the different combinations
of both treatments. A principal components analysis
(PCA) was carried out using the values of plastic
responses of the full set of plants to obtain new latent
variables reflecting the combination of responses that
best explained the variation observed in multi-trait
phenotypic plasticity. Then, mean values of the scores
September 2010 MULTI-TRAIT PHENOTYPIC PLASTICITY
of the first three principal components were calculated
for each treatment within each patch, and a two-way
ANOVA was used to assess whether competition and/
or herbivory promoted the expression of different
combinations of the induced responses observed. The
effects of competition and herbivory on plant fitness
were assessed using an ANOVA on the mean values of
the reproductive biomass produced by plants under
each combination of both treatments within each
patch. To meet the assumption of normality, such
values were ln-transformed.
To evaluate whether the ecological context changed
the targets of phenotypic selection and to assess whether
selection acted on independent plant responses or on
particular combinations of such responses, selection
analyses were performed separately for plants within
each treatment. Because the studied T. tubaeformis
population was structured in discrete patches or plots
within those patches (i.e., competition plots), replicates
were not completely independent and errors within each
patch were significantly correlated. Thus, contextual
analyses were used to calculate selection gradients, as
recommended by Heisler and Damuth (1987) for
structured populations. This approach allows the
identification of the impact of individual traits on plant
fitness, but also the influence of contextual characters,
defined as those that are shared identically by all
members of a particular group or patch. Such contextual
characters can be either aggregate characters resulting
from the combination of measurements made in
individuals of a given group (e.g., mean value of a trait
within a plot of a given patch) or characteristics
emerging as a property of each patch (e.g., plant density,
frequency of a given morph, etc.; Heisler and Damuth
1987). Thus, by simultaneously using lower and higher
level characteristics of individuals in a single analysis,
one can control for the intrinsic correlation between
individual characters and their group means and the
nonindependence of replicates (Heisler and Damuth
1987). Although commonly used to assess group
selection (Stevens et al. 1995, Tsuji 1995), this method
has been applied to study natural selection in structured
populations and metapopulations (Aspi et al. 2003). In
the case of this study, the means of each plant response
within each treatment and patch was calculated and
included in the selection analyses as contextual factors,
to account for the influence of plant aggregation on
relative fitness, through its effect on the mean value of
plant responses. Because the variation in the residuals of
these analyses was not explained by patch identity, these
factors are likely to account for most of the variation in
the influence of plant responses on plant fitness
associated with patch location. Relative fitness was
calculated using the mean fitness of plants within each
combination of both treatments, and its distribution
approached a Gaussian distribution. Standardized plant
responses within each treatment were used in general
linear models to calculate directional selection gradients
2631
at the individual and contextual levels. In a second set of
selection analyses using the contextual analysis ap-
proach, standardized trait values and their interactions
were used to calculate nonlinear and correlational
selection gradients (Lande and Arnold 1983), always
including the mean value of plant responses within each
patch as contextual variables. Correlative gradients
indicate that traits are being selected to become
significantly correlated (Phillips and Arnold 1989) or,
in other words, that particular combinations of traits are
being favored by selection. When more than two traits
are involved, however, correlative gradients are compli-
cated to interpret (Phillips and Arnold 1989). One
approach to understand high-order correlative gradients
is using selection analysis on the scores of a PCA with
the implicated traits. Although this method is tradition-
ally used to reduce the number of variables (Lande and
Arnold 1983), it can also be useful to assess how
selection acts on a particular combination of correlated
traits (Ordano et al. 2008), as each PC score in the
multidimentional space is related to a given value of
each trait. Thus, the covariance between PC scores and
relative fitness can be tracked down to assess the manner
in which the particular combination of correlated traits
are related to a given fitness value. However, this should
be an option only when a correlative selection gradient
has been previously detected with the traditional
selection analysis, to avoid misinterpretations of selec-
tion on PC scores (for a discussion, see Mitchell-Olds
1987). Hence, when significant correlative gradients
were detected using the multiple regression approach,
contextual analyses were also used within each ecolog-
ical scenario, this time using the scores of the principal
components and their means within each patch as
independent variables. Analyses were performed using
Statistica 7.1 (StatSoft 2005) or JMP 6.0 (SAS Institute
2006).
RESULTS
Significant effects of competition (Pillai’s trace ¼ 7.13,
F5,28 ¼ 39.95, P , 0.0001), but not of herbivory (Pillai’s
trace ¼ 0.21, F5,28 ¼ 1.17, P ¼ 0.34) were found on the
simultaneous expression of the five phenotypic respons-
es. In particular, competition during the juvenile stage
had negative impacts on the increase in diameter and
leaf production (Table 1, Fig. 1A). In contrast, stem
elongation and a greater internode length were promot-
ed by competition (Table 1, Fig. 1A). Realized resistance
was increased by competition and by early-season
damage (Table 1, Fig. 1B).
Plant responses were found to be correlated from low
to high degree (Table 2). Using PCA on such correlated
responses, three plant phenotypes were detected based
on the accumulated variance explained by the first three
PCs (85%): (1) PC1 defined vigorous plants, with wide
diameters and a large production of leaves, but with low
levels of realized resistance; (2) PC2 represented
elongated plants, with increased stem elongation and
2632 KARINA BOEGE Ecology, Vol. 91, No. 9

TABLE 1. F statistics of the univariate ANOVAs for each plant response under competition and/or
herbivory at the juvenile stage of the sunflower Tithonia tubaeformis.

Increase in Stem Internode Realized

Source diameter elongation length New leaves resistance
Competition 4.84* 10.69* 74.02**** 41.48**** 8.51***
Herbivory 0.005 0.07 0.06 0.001 4.71*
Competition 3 herbivory 0.76 0.18 1.68 0.61 0.71
Notes: For all traits, df ¼ 1, 32. Resistance is indirectly expressed by herbivore performance,
where high values indicate proportionally greater resistance than low values. This study was
conducted at the Pedregal San Ángel Reserve (PSAR), Mexico City, Mexico.
* P , 0.05; *** P , 0.001; **** P , 0.0001.

internode length but low numbers of new leaves and low resistance at the patch level (Table 4). Selection under
realized resistance; and (3) PC3 characterized resistant competition favored both diameter increase and stem
plants, with intermediate growth responses but a great elongation (Table 4). A marginally significant correla-
degree of realized resistance (Table 3). Univariate tive selection gradient was detected for the interaction
ANOVA on the mean values of PC scores within each between these two responses (c ¼ 0.71 (0.38), F1,63 ¼
treatment and patch detected that plants under compe- 3.73, P ¼ 0.06). Selection analyses on the PCs revealed
tition had low PC1 scores (F1,32 ¼ 24.25, P , 0.0001)
and a significant positive effect on PC2 (F1,32 ¼ 44.39,
P , 0.0001). Herbivory, in contrast, promoted plant
phenotypes with high scores of PC3 (F1,32 ¼ 6.15,
P ¼ 0.02).
Only competition had significant negative effects on
plant reproductive biomass. On average, plants with no
competition had 99.50 6 6.74 cm2 (mean 6 SE) of
reproductive biomass whereas plants under competition
had, on average, 14.82 6 1.72 cm2 (F1,32 ¼ 66.47, P ,
0.0001). In contrast, reproductive biomass was not
different between plants with (59.89 6 5.54 cm2) and
without early-season herbivory (54.04 6 5.55 cm2, F1,32
¼ 0.22, P ¼ 0.63).
The contextual analyses used to calculate selection
gradients revealed that the ecological context in which
plants grew modified the targets of phenotypic selection.
When plants grew alone without early-season herbivory,
a directional gradient was found for an increase in
diameter, favoring reduced realized resistance (Table 4).
A negative significant selection gradient was also found
at the patch level for the mean internode length (Table
4). This is, plants in patches that had environmental
conditions promoting plant elongation had lower
relative fitness than plants in other patches. In addition,
a correlative selection gradient (SE) was found for the
interaction among diameter increase 3 stem elongation
3 internode length 3 production of leaves (c ¼ 0.59
(0.30), F1,60 ¼ 3.87, P ¼ 0.05). For this second analysis, a
significant influence of the mean diameter increase per
patch on relative fitness was also found (b ¼ 0.75
(0.36), F1,60 ¼ 4.32, P ¼ 0.03). The selection analysis on
the scores of the PCs detected that in this environment, FIG. 1. Effects of herbivory and competition, alone and
together, on phenotypic plasticity of (A) stem diameter, stem
plants with high scores of PC1 and with low scores of elongation, internode length, and production of leaves and (B)
PC3 were favored (Table 5). The presence of herbivores realized resistance (mean 6 SE) in the wild sunflower Tithonia
during the early season modified these patterns of tubaeformis (Asteraceae) as indirectly expressed by herbivore
selection, given that only a negative directional selection performance, where high values indicate proportionally greater
resistance than low values. Different letters above bars
gradient was detected for internode length (Table 4) and represent significant differences among treatments (P , 0.05).
no correlative selection was detected. In addition, a This study was conducted at the Pedregal San Ángel Reserve
negative selection gradient was detected for realized (PSAR), Mexico City, Mexico.
September 2010 MULTI-TRAIT PHENOTYPIC PLASTICITY 2633

TABLE 2. Pairwise Pearson r correlations among the five plastic responses in Tithonia tubaeformis.

Increase in Stem Internode New Realized

Response diameter elongation length leaves resistance
Increase in diameter 1 0.47**** 0.26**** 0.61**** 0.23****
Stem elongation 1 0.18**** 0.09* 0.01
Internode length 1 0.44**** 0.12**
New leaves 1 0.23****
Realized resistance 1
* P , 0.05; ** P , 0.01; **** P , 0.0001.

that plants with high scores of PC1, PC2, and PC3 were
favored. Finally, when both competition and herbivory
occurred together, they favored plants with the greatest
diameter increase and production of leaves (Table 4). In
addition, in this ecological environment, a significant
correlative selection was detected on the interaction
among diameter increase 3 internode length 3 stem
elongation 3 production of leaves 3 realized resistance
(c ¼ 2.54 (1.61), F1,47 ¼ 4.73, P ¼ 0.03). Accordingly,
selection was detected for PC1 and PC3, favoring plants
with high scores of both PCs (Table 5). Interestingly,
in both environments with competition a negative
selection gradient on PC3 at the patch level was detected
(Table 5).
DISCUSSION
The results of this study suggest that plants have the
ability to respond with particular combinations of
plastic traits, depending on the ecological context in
which they grow. More interestingly, they also show that
natural selection acts on particular combinations of
traits, depending on the presence of competitors and/or
herbivores. In general, in most environments, natural
selection favored the combination of traits that provided
plant vigor (PC1), whereas realized resistance (PC3) was
favored only when both interactions occurred simulta-
neously. Selection also favored the combination of traits
that allowed plant elongation (PC2) when plants grew
under competition. Overall, the results from this study
suggest that natural selection is likely to act simulta-
neously on multiple correlated plastic traits, depending
on the community context, and that environmental
heterogeneity can affect the overall results of natural
selection.
Interestingly, competition alone increased realized
resistance of plants. Such a decrease in plant quality is
likely to correspond to an increase in constitutive
resistance as a consequence of the accumulation of
carbon-based secondary metabolites (e.g., terpenes) if
plant growth is limited by neighbors (Stamp 2003). In
particular, this plastic response has been previously
reported for the constitutive production of terpenes,
compounds produced by other species of Tithonia
(Ambrosio et al. 2008), as a result of changes in nutrient
availability (Mihaliak and Lincoln 1989). Changes in
nitrogen and water contents are unlikely to have been
responsible for such increased resistance given that no
correlations have been found between traits and
herbivore performance (K. Boege and A. Hernández,
unpublished data). In addition, neighboring plants also
promoted a multi-trait plastic response characterized by
increased stem elongation and internode length, together
with a reduction in stem diameter and production of
new leaves (high values of PC2). Stem elongation has
been reported in other systems to be an adaptive plastic
trait (Schmitt et al. 1995, Dudley and Schmitt 1996,
Dorn et al. 2000) when light is limiting under crowded
conditions (Lee and Bazzaz 1980, Ballaré et al. 1990,
Callaway et al. 2003). For T. tubaeformis, this trait also
seems to be an expression of adaptive phenotypic
plasticity, but only when expressed with a particular
combination of the other traits (i.e., PC2), because it
was favored by natural selection in the environment in
which it was expressed (competition; Dorn et al. 2000).
In contrast with other systems, such as in Chenopodium
albus (Kurashige and Agrawal 2005), the elongation
response in T. tubaeformis was not affected by
herbivory.
Increased resistance was found in plants that received
early-season damage, and there was a tendency of this
response to be even greater when herbivory occurred
simultaneously with competition. This tendency suggests
an increase in both inducible and constitutive resistance
traits. Recent evidence has demonstrated that the costs
of induced resistance in Arabidopsis thaliana are greater
in high-nutrient, noncompetitive environments than in
low-nutrient, competitive environments (Cipollini 2007,
2010). This could be a possible mechanism to explain
why resistance induced by herbivores tended to be
TABLE 3. Eigenvalues and eigenvectors of first three principal
components, which explained 85% of the variance of
diameter increase, stem elongation, internode length, pro-
duction of new leaves, and realized resistance across all
treatments.
Principal component PC1 PC2 PC3
Eigenvalue 2.06 1.28 0.91
Percentage 41.38 25.72 18.12
Eigenvectors
Diameter increase 0.60 0.27 0.09
Internode length 0.39 0.55 0.26
Stem elongation 0.22 0.77 0.05
New leaves 0.59 0.14 0.16
Realized resistance 0.29 0.10 0.95
2634 KARINA BOEGE Ecology, Vol. 91, No. 9

TABLE 4. Directional selection gradients (with SE in parentheses) from contextual analysis on phenotypic and contextual
characters (their mean values per patch) under different ecological scenarios.

Increase in Internode Stem Realized

Directional gradients df diameter length elongation New leaves resistance
Trait effects on fitness
Control 1, 91 0.28 (0.14)* NS NS NS 0.21(0.08)**
Herbivory 1, 73 NS 0.37(0.12)** NS NS NS
Competition 1, 94 0.73 (0.15)**** NS 0.38 (0.17)* NS NS
Competition þ herbivory 1, 78 0.79 (0.21)**** NS NS 0.67 (0.17)**** NS
Contextual effects on fitness
Control 1, 91 NS 0.46(0.41)* NS NS NS
Herbivory 1, 73 NS NS NS NS 0.78*
Competition 1, 94 NS NS NS NS NS
Competition þ herbivory 1, 78 NS NS NS NS NS
Notes: Directional selection gradients were obtained from a selection analysis using a contextual analysis approach, which
included all five induced responses and their means per patch as independent variables in a multiple regression on relative fitness
values. For more details see Methods: Statistical analyses.
* P , 0.05; ** P , 0.01; **** P , 0.0001; NS, nonsignificant.

greater when plants were competing with neighbors.
Furthermore, regardless of the costs of defense, induced
resistance could have been improved by plant competi-
tion through the influence of stress-regulating hormones
on the metabolic pathways regulating the induction of
resistance to herbivores. For example, abscisic acid,
which is an important hormone regulator of responses
to water and salt stress, has been found to enhance the
jasmonate-induced defenses against herbivores in toma-
to plants (Thaler and Bostock 2004). Moreover,
increased resistance, and/or other functions sharing
metabolic pathways with such response, can provide
some benefits to plants growing under competition
(Cipollini 2004, Jones et al. 2006). For example, the
signaling role of jasmonic acid induced after herbivore
attack has been reported to be associated with responses
of plants to water (Reymond et al. 2000) and salt (Wang
et al. 2001) stress. Thus, by being activated by
herbivores, signaling molecules could enhance responses
to plant competition. In addition, secondary metabolites
induced by herbivores could also have allelopathic
properties (Inderjit 1996, Lankau and Strauss 2007),
reducing the costs of defense under competitive envi-
ronments, as has been found for Arabis perennans
(Siemens et al. 2003) and A. drummondii (Jones et al.
2006). The benefits of an increased realized resistance in
the competitive environment was also suggested by the
positive selection gradient observed for this response.
In accordance with the Active Response Model
(Karban et al. 1989), several studies have found that
induced responses by competition actually constrain
plant responses to herbivores (Bergelson 1994, Cipollini
and Bergelson 2001, Tiffin 2002). The present study,
however, shows the opposite trend and rather adds to
the evidence supporting the recently proposed Defense
Stress Benefit Hypothesis (Siemens et al. 2003), suggest-
ing that, in fact, resistance can be enhanced in
competitive environments. Overall, the accumulation
of studies supporting both hypotheses suggests that the
interactive effects of competition and herbivory on plant
resistance may be driven not only by resource trade-offs,
but by more complex dynamics given the putative
multiple functions of such responses (Jones et al. 2006)
and may also depend on which resources are depleted by
competition (Agrawal 2004, Wise and Abrahamson
2005) and their influence on both constitutive and
induced resistance traits.

TABLE 5. Directional selection gradients (with SE in parentheses) from contextual analysis on principal components (PC) scores and
the corresponding contextual traits (their mean values per patch) under different ecological scenarios.

Directional gradient df PC1 PC2 PC3

Effects on fitness
Control 1, 95 0.44 (0.09)**** NS 0.31 (0.16)*
Herbivory 1, 79 0.58 (0.12)**** NS NS
Competition 1, 98 0.85 (0.10)**** 0.42 (0.09)**** 0.19 (0.04)*
Competition þ herbivory 1, 82 0.90 (0.13)**** NS 0.28 (0.09)**
Contextual effects on fitness
Control 1, 95 NS NS NS
Herbivory 1, 79 NS NS 1.11 (0.38)**
Competition 1, 98 NS NS 0.60 (0.30)*
Competition þ herbivory 1, 82 0.75 (0.22)** NS 0.43 (0.18)*

* P , 0.05; ** P , 0.01; **** P , 0.0001; NS, nonsignificant.

September 2010 MULTI-TRAIT PHENOTYPIC PLASTICITY
In contrast with studies that have found negative
impacts of damage at the juvenile stage on plant fitness
(Ehrlén 1995, Hanley et al. 1995, Warner and Cushman
2002, Boege et al. 2007) and described how such impacts
increase under crowded conditions (Tiffin 2002), early-
season damage on T. tubaeformis had no evident impact
on plant fitness. These findings suggest that T. tubae-
formis plants were able to completely tolerate damage at
juvenile stages, probably because they had enough time
to recover for the tissue lost during their juvenile stage.
Alternatively, damage accumulated during that period
may not have been enough to produce significant
impacts on plant fitness. Accumulation of damage
during later stages of the plants’ life cycle is likely to
be more harmful for T. tubaeformis, as has been found
for other species (Warner and Cushman 2002).
Although commonly used (Schluter and Smith 1986,
Campbell 1989, Parra-Tabla and Bullock 2000), selec-
tion analyses on PCs have been criticized, given the risk
of misleading conclusions of the targets of selection
(Mitchell-Olds and Shaw 1987). If significant selection is
observed for individual traits with high loadings on a
PC, selection on that PC is likely to be driven by the
independent trait and not by the particular combination
of traits defining the PC (Mitchell-Olds and Shaw 1987).
However, this approach can be useful when used to
interpret selection on particular combinations of traits in
cases in which significant high-order correlational
gradients have been previously identified with tradition-
al selection analyses, which was the case in this study.
Selection analyses detected that the ecological context in
which plants grew modified the selective pressures on
different plastic responses, and more interestingly, on
particular combinations of their expression. In the
absence of competition and early-season herbivory,
selection favored plants with increased diameter, high
production of leaves, with a mild stem elongation but
short internodes (high scores of PC1), and with low
realized resistance (low scores of PC3). The correlative
structure among traits did not seem to impose any
restriction for the selection on this particular combina-
tion of traits. The fact that selection favored plants with
low levels of realized (constitutive) resistance in this
environment supports the notion that resistance is costly
(Baldwin and Hamilton 2000) and that costs of
resistance are greater when plants have high resource
availability (Cipollini 2010). Contextual-level selection
against plants in patches that promoted large internode
lengths suggests that plant elongation can be costly
when no competitors are present (Cipollini and Schultz
1999), probably because it reduces the possibility of
plants to produce vigorous phenotypes. In the environ-
ment with early-season herbivory, patterns of selection
changed, given that, within each patch, elongated plants
with large internode lengths were not favored by
selection. This could be related to the expression of
tolerance in damaged plants. Because plants with
herbivory were thinner, shorter, and had fewer leaves,
2635
a reduction in internode length and stem elongation
could have allowed plants to produce more leaves and
replace a greater amount of tissue lost to herbivory early
in the season. Interestingly, in this environment, patch-
level selection was found against realized resistance,
supporting the notion that tolerance is likely to be a
more effective mechanism to deal with early-season
herbivory than induced resistance. Thus, patches with
plants with increased realized resistance probably have
fewer resources for the expression of tolerance, if a
trade-off prevails between these two defensive functions
(Stowe et al. 2000).
In both environments with competition, plants with
increased diameter were favored by natural selection.
This selective advantage was probably due to an
increased ability to structurally support more photosyn-
thetic tissue of elongated plants, which in general were
very prone to fall down. Furthermore, the significant
selection gradients for PC1 and PC2 suggest that plant
elongation was also favored, although only in the
absence of herbivores. This could be, again, a conse-
quence of the negative impact of plant elongation on the
ability of plants to compensate tissue lost to herbivores.
Notice that because competition produced phenotypes
with low values of PC1 and high values of PC2, any
plant with intermediate values of PC1 and PC2 should
have had a selective advantage. Interestingly, although
realized resistance was favored at the individual level
within each patch in both environments under compe-
tition, unknown factors related to patch location
promoted lower fitness in patches where plants, on
average, showed greater realized resistance. It is likely
that those factors had to do with plant vigor and
resource allocation trade-offs between resistance and
other functions, but this warrants further investigation.
Overall, the results from this study suggest that the
presence of a third species (e.g., competitors) can modify
the selective impact of one species (e.g., herbivores) on
another (e.g., plants), as has been found in previous
studies (Agrawal and Van Zandt 2003, Strauss and
Irwin 2004). Moreover, the use of contextual analyses
revealed that such selective impact is also affected by
population structure and environmental heterogeneity.
Differences in the targets of natural selection as a
function of the ecological and environmental context
highlight the relevance of the expression of phenotypic
plasticity. Natural selection acted on independent traits
for some plastic responses, but on combinations of traits
for other responses. This may allow plants to adjust
their phenotypes to those that promote greater fitness
under particular ecological conditions.
ACKNOWLEDGMENTS
I thank R. Pérez, S.Ochoa, A. Mendoza, and X. Damian for
their invaluable help in the field. Special thanks to S. Benitez-
Vieyra for help with statistical analyses. R. J. Fornoni, D.
Siemens, A. Agrawal, M. Johnson, P. Tiffin, E. del Val, D.
Carmona, E. Garrido, M. Abarca, A. Hernandez, and two
anonymous reviewers kindly reviewed earlier versions of the
2636 KARINA BOEGE
manuscript, which was improved by their valuable comments.
This project was funded by PAPIIT-UNAM (IN-225007).
LITERATURE CITED
Adler, L. S., M. Wink, M. Distl, and A. J. Lentz. 2006. Leaf
herbivory and nutrients increase nectar alkaloids. Ecology
Letters 2006:960–967.
Agrawal, A. A. 1998. Induced responses to herbivory and
increased plant performance. Science 279:1201–1202.
Agrawal, A. A. 2004. Resistance and susceptibility of milkweed:
competition, root herbivory, and plant genetic variation.
Ecology 85:2118–2133.
Agrawal, A. A., J. A. Lau, and P. A. Hambäck. 2006.
Community heterogeneity and the evolution of interactions
between plants and insect herbivores. Quarterly Review of
Biology 81:349–376.
Agrawal, A. A., and P. Van Zandt. 2003. Ecological play in the
coevolutionary theatre: genetic and environmental determi-
nants of attack by a specialist weevil on milkweed. Journal of
Ecology 91:1049–1059.
Ambrosio, S. R., Y. Oki, V. C. Gomes Heleno, J. S. Chaves,
P. G. B. D. Nascimento, J. E. Lichston, M. G. Constantino,
E. M. Varanda, and F. B. Da Costa. 2008. Constituents of
glandular trichomes of Thitonia diversifolia: relationships to
herbivory and antifeedant activity. Phytochemistry 69:2052–
2060.
Aspi, J., A. Jäkäläniemi, J. Tuomi, and P. Sikamäki. 2003.
Multilevel phenotypic selection on morphological characters
in a metapopulation of Silene tatarica. Evolution 57:509–517.
Baldwin, I. T. 1998. Jasmonate-induced responses are costly but
benefit plants under attack in native populations. Proceed-
ings of the National Academy of Sciences USA 95:8113–
8118.
Baldwin, I. T., and J. G. Hamilton. 2000. Jasmonate-induced
responses of Nicotiana sylvestris are costly due to impaired
competitive ability for nitrogen. Journal of Chemical Ecology
26:915–952.
Ballaré, C. L., A. L. Scopel, and R. A. Sánchez. 1990. Far-red
radiation reflected from adjacent leaves: early signal of
competition in plant canopies. Science 247:329–332.
Bergelson, J. M. 1994. The effects of genotype and the
environment on costs of resistance in lettuce. American
Naturalist 143:349–359.
Boege, K., and R. Dirzo. 2004. Intraspecific variation in
growth, defense and herbivory in Dialium guianense (Caes-
alpiniaceae) mediated by edaphic heterogeneity. Plant
Ecology 175:59–69.
Boege, K., R. Dirzo, D. H. Siemens, and P. Brown. 2007.
Ontogenetic switches from plant resistance to tolerance:
Minimizing costs with age? Ecology Letters 10:177–187.
Callaway, R. M., S. C. Pennings, and C. L. Richards. 2003.
Phenotypic plasticity and interactions among plants. Ecology
84:1115–1128.
Campbell, D. R. 1989. Measurements of selection in a
hermaphroditic plant: variation in male and female pollina-
tion success. Evolution 43:318–334.
Cipollini, D. F. 2002. Does competition magnify the fitness
costs of induced responses in Arabidopsis thaliana? A
manipulative approach. Oecologia 131:514–520.
Cipollini, D. 2004. Stretching the limits of plasticity: Can a
plant defend against both competitors and herbivores?
Ecology 85:28–37.
Cipollini, D. F. 2007. Consequences of the overproduction of
methyljasmonate on seed production, tolerance to defoliation
and competitive effect and response of Arabidopsis thaliana.
New Phytologist 173:146–453.
Cipollini, D. F. 2010. Constitutive expression of methyl
jasmonate-inducible responses delays reproduction and
constrains fitness responses to nutrients in Arabidopsis
thaliana. Evolutionary Ecology 24:59–68.
Ecology, Vol. 91, No. 9
Cipollini, D., and J. Bergelson. 2001. Plant density and nutrient
availability constrain the constitutive wound-induced expres-
sion of tripsin inhibitors in Brassica napus. Journal of
Chemical Ecology 27:593–610.
Cipollini, D. F., and J. C. Schultz. 1999. Exploring cost
constraints on stem elongation in plants using phenotypic
manipulation. American Naturalist 153:236–242.
Crawley, M. J. 1997. Plant herbivore dynamics. Pages 401–475
in M. J. Crawley, editor. Plant ecology. Blackwell Science,
Cambridge, UK.
Dorn, L. A., E. H. Pyle, and J. Schmitt. 2000. Plasticity to light
cues and resources in Arabidopsis thaliana: testing for
adaptive value and costs. Evolution 54:1982–1994.
Dudley, S. A., and J. Schmitt. 1996. Testing the adaptive
plasticity hypothesis: density-dependent selection on manip-
ulated stem length in Impatiens capensis. American Naturalist
147:445–465.
Ehrlén, J. 1995. Demography of the perennial herb Lathyrus
vernus. I. Herbivory and individual performance. Journal of
Ecology 83:287–295.
Guerevich, J., J. A. Morrison, and L. V. Hedges. 2000. The
interaction between competition and predation: a meta-
analysis of field experiments. American Naturalist 155:435–
453.
Hambäck, P. A., and A. P. Beckerman. 2003. Herbivory and
plant resource competition: a review of two interacting
interactions. Oikos 101:26–37.
Hanley, M. E., M. Fenner, and P. J. Edwards. 1995. The effect
of seedling age on the likelihood of herbivory by the slug
Deroceras reticulatum. Functional Ecology 9:754–759.
Heisler, I. L., and J. Damuth. 1987. A method for analyzing
selection in hierarchically structured populations. American
Naturalist 130:582–602.
Herms, D. A., and W. J. Mattson. 1992. The dilemma of plants:
to grow or to defend. Quarterly Review of Biology 62:186–
202.
Inderjit. 1996. Plant phenolics in allelopathy. Botanical Review
62:186–202.
Inderjit, and R. Del Moral, 1997. Is separating resource
competition from allelopathy realistic? Botanical Review 63:
221–230.
Izaguirre, M. M., C. A. Mazza, M. Biondini, I. T. Baldwin, and
C. L. Ballaré. 2006. Remote sensing of future competitors:
impacts on plant defense. Proceedings of the National
Academy of Sciences USA 103:7170–7174.
Jones, T. H., S. Kulseth, K. Mechtenberg, C. Jorgenson, M.
Zwhfya, P. Brown, and D. H. Siemens. 2006. Simultaneous
evolution of competitiveness and defense: induced switching
in Arabis drummondii. Plant Ecology 184:245–257.
Karban, R., A. A. Agrawal, J. S. Thaler, and L. S. Adler. 1999.
Induced plant responses and information content about risk
of herbivory. Trends in Ecology and Evolution 14:443–447.
Karban, R., and I. T. Baldwin. 1997. Induced responses to
herbivory. University of Chicago Press, Chicago, Illinois,
USA.
Karban, R., A. K. Brody, and W. C. Schnathorst. 1989.
Crowding and a plant’s ability to defend itself against
herbivores and diseases. American Naturalist 134:749–760.
Kurashige, N. S., and A. A. Agrawal. 2005. Phenotypic
plasticity to light competition and herbivory in Chenopodium
album (Chenopodiaceae). American Journal of Botany 92:
21–26.
Lande, R., and S. J. Arnold. 1983. The measurement of
selection on correlated characters. Evolution 37:1210–1226.
Lankau, R. A., and S. Y. Strauss. 2007. Mutual feedbacks
maintain both genetic and species diversity in a plant
community. Science 317:1561–1563.
Lee, T. D., and E. A. Bazzaz. 1980. Effects of defoliation and
competition on growth and reproduction in the annual plant
Abutilon theophrasti. Journal of Ecology 68:813–821.
September 2010 MULTI-TRAIT PHENOTYPIC PLASTICITY
Maschinski, J., and T. G. Whitham. 1989. The continuum of
plant responses to herbivory: the influence of plant associ-
ation, nutrient availability and timing. American Naturalist
134:1–19.
McGuire, R., and A. A. Agrawal. 2005. Trade-offs between the
shade-avoidance response and plant resistance to herbivores?
Tests with mutant Cucumis sativus. Functional Ecology 19:
1025–1031.
McVaugh, R. 1984. Compositae. Flora Novo-Galiciana. A
descriptive account of the vascular plants of western Mexico.
University of Michigan Press, Ann Arbor, Michigan, USA.
Mihaliak, C. A., and D. E. Lincoln. 1989. Plant biomass
partitioning and chemical defense: response to defoliation
and nitrate limitation. Oecologia 80:122–126.
Mitchell-Olds, T., and R. G. Shaw. 1987. Analysis of natural
selection: statistical inference and biological interpretation.
Evolution 41:1149–1161.
Ordano, M., J. Fornoni, K. Boege, and C. Domı́nguez. 2008.
The adaptive value of phenotypic floral integration. New
Phytologist 179:1183–1192.
Parra-Tabla, V., and S. H. Bullock. 2000. Phenotypic natural
selection on flower biomass allocation in the tropical tree
Ipomoea wolcottiana Rose (Convolvulaceae). Plant System-
atics and Evolution 221:167–177.
Phillips, P. C., and S. J. Arnold. 1989. Visualizing multivariate
selection. Evolution 43:1209–1222.
Reymond, P. H., H. Weber, M. Damond, and E. E. Farmer.
2000. Differential gene expression in response to mechanical
wounding and insect feeding in Arabidopsis. Plant Cell 12:
707–719.
Rzedowski, J. 1954. Vegetación del Pedregal de San Ángel.
Anales de la Escuela Nacional de Ciencias Biológicas 8:59–
129.
SAS Institute. 2006. JMP. Release 6.0.3. Statistical discovery.
SAS Institute, Cary, North Carolina, USA.
Schluter, D., and J. N. M. Smith. 1986. Natural selection on
beak and body size in the song sparrow. Evolution 40:221–
231.
Schmitt, J., A. C. McCormac, and H. Smith. 1995. A test of the
adaptive plasticity hypothesis using transgenic and mutant
plants disabled in phytochrome-mediated elongation re-
sponses to neighbors. American Naturalist 146:937–953.
Siemens, D. H., S. H. Garner, T. Mitchel-Olds, and R. M.
Callaway. 2002. Cost of defense in the context of plant
competition: Brassica rapa may grow and defend. Ecology 83:
505–517.
Siemens, D. H., H. Lischke, N. Maggiulli, S. Schurch, and B. A.
Roy. 2003. Cost of resistance and tolerance under competi-
2637
tion: the defense-stress benefit hypothesis. Evolutionary
Ecology 17:247–263.
Stamp, N. 2003. Out of the quagmire of plant defense
hypotheses. Quarterly Review of Biology 78:23–55.
StatSoft. 2005. Statistica. Version 7.1. StatSoft, Tulsa, Okla-
homa, USA.
Stevens, L., C. J. Goodnight, and S. Kalisz. 1995. Multilevel
selection in natural populations of Impatiens capensis.
American Naturalist 145:513–526.
Stinchcombe, J. R., and M. D. Rausher. 2001. The evolution of
tolerance to deer herbivory: modifications caused by the
abundance of insect herbivores. Proceedings of the Royal
Society B 269:1241–1246.
Stowe, K. A., R. J. Marquis, C. G. Hochwender, and E. L.
Simms. 2000. The evolutionary ecology of tolerance to
consumer damage. Annual Review of Ecology and System-
atics 31:565–595.
Strauss, S. Y., and R. E. Irwin. 2004. Ecological and
evolutionary consequences of multispecies plant–animal
interactions. Annual Review of Ecology and Systematics
35:435–466.
Strauss, S. Y., D. H. Siemens, M. B. Decher, and T. Mitchell-
Olds. 1999. Ecological costs of plant resistance to herbivores
in the currency of pollination. Evolution 53:1105–1113.
Thaler, J., and R. M. Bostock. 2004. Interactions between
abscisic acid-mediated responses and plant resistance to
pathogens and insects. Ecology 85:48–58.
Tiffin, P. 2002. Competition and time of damage affect the
pattern of selection acting on plant defense against herbi-
vores. Ecology 83:1981–1990.
Tsuji, K. 1995. Reproductive conflicts and levels of selection in
the ant Pristomyrmex pungens: contextual analysis and
partitioning of covariance. American Naturalist 146:586–
607.
Wang, Y., S. Mopper, and K. H. Hasenstein. 2001. Effects of
salinity on endogenous ABA, IAA, JA and SA in Iris
hexagona. Journal of Chemical Ecology 27:327–342.
Warner, P. J., and J. H. Cushman. 2002. Influence of herbivores
on a perennial plant: variation with life history stage and
herbivore species. Oecologia 132:77–85.
Weis, A. E., and M. E. Hochberg. 2000. The diverse effects of
intraspecific competition on the selective advantage to
resistance: a model and its predictions. American Naturalist
156:276–292.
Wise, M. J., and W. G. Abrahamson. 2005. Beyond the
compensatory continuum: environmental resource levels and
plant tolerance of herbivory. Oikos 109:417–428.