vol. 167, no. 1 the american naturalistjanuary 2006
Alternative Designs and the Evolution of Functional Diversity
Christian O. Marks*and Martin J. Lechowicz†
Biology Department, McGill University, 1205 Dr. Penfield Avenue,
Montreal, Quebec H3A 1B1, Canada
Submitted February 28, 2005; Accepted August 5, 2005;
Electronically published October 11, 2005
Online enhancement: table.
abstract: According to conventional wisdom, functional diversity
is exclusively a consequence of species having evolved adaptations
to fill different niches within a heterogeneousenvironment.Thisview
anticipates only one optimal combination of trait values in a given
environment, but it is also conceivable that alternative designs of
equal fitness in the same environment might evolve. To investigate
that possibility, we use a genetic algorithm to search for optimal
combinations of 34 functional traits in a realistic model of tree seed-
ling growth and survival. We show that separate lineages of seedlings
evolving in identical environments result in many alternative func-
tional designs of approximately equal fitness.
Keywords: alternative designs, convergent evolution, functional di-
versity, niche overlap, trait adaptation, tree seedlings.
One of the most striking observations from tropical low-
land rain forests to boreal conifer forests is the high levels
of quantitative variation in traits among co-occurring tree
species. The within-site variation among species in the
value of traits critical to plant function usually exceeds an
order of magnitude (Niinemets 2001; Maherali et al. 2004;
Wright et al. 2004) and can be greater than among-site
variation (Westoby et al. 2002). This observation raises the
question of why the values of traits have not converged
to a greater degree in species that are currently growing
in the same environment (Ko ¨rner 1991). Here we show
that in evolution, a set of traits can be combined in quan-
titatively different ways to achieve approximately equal
fitness in a given environment, which can account for the
observed functional diversity among co-occurring trees in
* E-mail: firstname.lastname@example.org.
Am. Nat. 2006. Vol. 167, pp. 55–66. ? 2006 by The University of Chicago.
0003-0147/2006/16701-40915$15.00. All rights reserved.
contemporary forests. Furthermore, we show that this
mechanism is a general one that could contribute to the
evolution of functional diversity in all types of organisms.
The traditional view of trait evolution holds that selec-
tion favors individuals well suited to their environment,
individuals with traits and trait combinations that increase
fitness. For tree seedlings, fitness can be assumed to be
the combination of survival and maximization of growth.
Maximizing seedling growth increases the chance to sur-
vive into adulthood and increases lifetime reproductive
potential (Van Valen 1975; Harcombe 1987; Oliver and
Larson 1996). Growth and survival are in turn a function
of the resource balances of the seedling, particularly light
(Pacala et al. 1994), carbon (Kobe 1997), water (Caspersen
and Kobe 2001), and nutrients (Aerts and Chapin 2000).
Traits should evolve toward values that improve these re-
source balances and thus growth rate in a given environ-
ment (Parker and Maynard Smith 1990; Ma ¨kela ¨ et al.
This view has been formalized in various mathematical
optimization models (Parker and Maynard Smith 1990;
Ma ¨kela ¨ et al. 2002; Sutherland 2005). Typically, thesemod-
els focus on a single trade-off that governs a particular
trait of interest. We refer to this class of models as single
trade-off models. A trade-off is described by both costs
and benefits that vary in some nonlinear fashion with the
value of the trait. Consequently, there is usually a single
optimal trait value that maximizes fitness. If the equations
modeling these costs and benefits are cleverly chosen, then
it is often possible to solve for the optimal trait value
analytically, which has made these models enticing. We
will not elaborate an explanation of single trade-off mod-
els; the approach is reviewed elsewhere (Parker and May-
nard Smith 1990; Ma ¨kela ¨ et al. 2002). The key point here
is the contradictory observation that single trade-off mod-
els predict a single optimal trait value in a particular en-
vironment, yet field researchers consistently measure great
variation in trait values among tree species at a given site
(Niinemets 2001; Westoby et al. 2002; Maherali et al. 2004;
Wright et al. 2004).
How can the prediction of the single optimal trait value
be reconciled with the great variation observed in the field?
The standard ecological explanation is that the field site
56 The American Naturalist
must be heterogeneous (in either time or space) and thus
does not represent a single environment but rather a com-
plex of microenvironments or a series of fluctuating en-
vironments, each selecting for different optimaltraitvalues
(Levin and Muller-Landau 2000). These explanations in-
clude not only the abiotic environmentalheterogeneitybut
also the environmental heterogeneity imposed by the or-
ganisms occupying the environment (e.g., Iwasa et al.
1984). However, it seems doubtful to us that the relatively
minor environmental variation within a site could explain
all of the large trait variation observed in co-occurring
plant species, especially given the typically low correlation
between trait values and environmental variables (Niine-
mets 2001; Maherali et al. 2004; Wright et al. 2004).
The propensity of trees to be habitat generalists casts
further doubt on local heterogeneity as the explanation
for functional diversity at a site. For example, ecologists
are discovering that in the diverse forests of thewettropics,
most tree species are habitat generalists (Ricklefs 2000;
Pitman et al. 2001; Valencia et al. 2004). This implies fre-
quent niche overlap in these species despite great trait
diversity. Even in the less diverse temperate forests, many
species are generalists (Burns and Honkala 1990), and in-
depth studies of tree species distributions on environ-
mental gradients have shown large overlap in niches
(Whittaker 1956; Fralish et al. 1978; Iverson et al. 1999;
Cavender-Bares et al. 2004a). In fact, this niche overlap is
so great that it has led some researchers to propose that
niche differences can be neglected in neutral models of
community assembly (Bell 2000; Hubbell 2001, 2005).
We recognize the importance of environmental gradi-
ents and niche partitioning in promoting diversity, espe-
cially in the radiation of a single lineage (Schluter 2000;
Cavender-Bares et al. 2004b). However, the ecological pat-
terns discussed above cast substantial doubt on environ-
mental heterogeneity as the sole explanation for the var-
iation in plant traits encountered within a locality for
species from diverse lineages. As an alternative, wepropose
that if a model of tree seedling growth and survival were
based on the complex interactions among multiple trade-
offs instead of a single trade-off, then multiple optima
could emerge even in a single environment. Multiple op-
tima imply that diverse trait combinations are feasible in
the same habitat, each representing an alternative func-
tional design of approximately equal competence. Because
of the effects of history and chance, different lineages
would be expectedto evolvetowarddifferentoptima(Bock
1959, 1976, 1980; Lewontin 1978; Gould 1989). For ex-
ample, it is conceivable that a red oak–type design and a
white oak–type design may be equally well adapted to the
same habitat despite substantial trait differences (Caven-
der-Bares et al. 2004b).
We have a high regard for the elegance of single trade-
off models and the interesting insights that can be gained
from such models. However, it is important to recognize
that single trade-off models isolate some part of plant
function and thereby limit the possibilities for adaptation
at the whole-plant level (Gould and Lewontin 1979; Mayr
1983). In real organisms and especially in plants, traits
often have a critical role in more than a single function,
and trait values therefore reflect the outcome of trade-offs
among functions that are not readily assessed in a single
trade-off model. More holistic models that explicitly con-
sider the many trait interactions governing function allow
us to better explore the consequences of this complexity
at the whole-plant level (Ellis 2005; Proulx et al. 2005).
Specifically, in this article, we take a holistic approach to
test the hypothesis that the complex interactions among
multiple traits result in the existence of alternative, more
or less equally competent functional designs in a given
environment. We present the existence of alternative,equi-
competent functional designs as a likely mechanism that
can account for the high levels of trait variation in tree
species with similar habitat affinities. Our simulation re-
sults provide a novel explanation of how the functional
diversity among tree species observed in a locality might
have evolved and offer some indication of the degree of
functional redundancy inherent in species diversity.
To test our hypothesis, we developed a simulation model
of tree seedling growth and survival combined with a ge-
netic algorithm (GA) to search for optimal trait combi-
nations. We describe this tree seedling adaptive design
model (TAD) in detail elsewhere (Marks and Lechowicz,
forthcoming) but outline the model here to support our
consideration of alternative functional designs.
Seedling Simulation Model
The TAD model is based on the interactions and partic-
ularly the trade-offs among 34 key functional traits. The
resources considered in the modeled trade-offs include
light, carbon, water, and nitrogen. We refer to the 34 trait
variables that are subject to optimization in the model as
independent traits, whereas other traits that are derived
from these independent traits are referred to as dependent
traits. Consideration of the trade-offs among this relatively
large number of traits has several advantages. First, it al-
lows testing the hypothesis that multipleoptimaforagiven
environment can result from the complex interactions
among multiple traits. Second, interactions among the
traits can be integrated at the whole-seedling level to give
a single realistic fitness measure (i.e., survival and growth
rate). Finally, the rich array of interacting traits increases
Alternative Functional Designs57
the realism of the simulated seedlings and lets the key
traits and derived traits be compared with reported data
for natural seedlings in diverse environments.
The 34 traits were selected after a thorough review of
the literature on the functional ecology of woody plants.
The primary criterion for letting a trait be an independent
variable in the model was that the trait has a substantial
and well enough understood effect on one or more trade-
offs involving resource balances to model accurately. Be-
cause leaves have been studied more intensively thanroots,
the model has a more detailed representation of above-
ground than belowground traits, but many belowground
traits andprocesses areincluded. The34independenttraits
include four parameters related to seed reserve allocation,
six parameters related to carbon allocation, three param-
eters involved in nitrogen allocation, three parameters for
stomatal control, nine leaf traits, five root traits, and four
The essence of the TAD model is the ensemble of func-
tional trade-offs and their interactions. We introduce the
modeled trade-offs in the following paragraphs to give the
reader some insight into their complexity. We present else-
where a more detailed description and validation of the
model, including equations defining the functional trade-
offs and trait interactions together with a supporting lit-
erature review (Marks and Lechowicz, forthcoming). The
brief summary of the model here is intended to be didactic
and omits supporting references and details to emphasize
the overall nature of the TAD model.
Leaves use all four modeled resources (light, carbon,
water, and nitrogen), making the interactions among leaf
traits particularly complex. For example, leaves have sto-
matal pores to regulate gas exchange with the surrounding
air. Increasing stomatal conductance allows more CO2to
diffuse into the leaf, thereby increasing the photosynthetic
rate, but at the same time increases the amount of water
lost from the leaf through transpiration. Becausesupplying
the leaf with water is costly, particularly in dry environ-
ments, a plant should regulate its stomatal conductance
such that it maximizes photosynthesis relative to transpi-
ration. Similarly, leaves have a cuticle to reduce water
losses. Investing in a thicker cuticle improves survival dur-
ing a prolonged drought but also increases leaf construc-
tion cost. Thus, there are multiple trade-offs between the
carbon and water economy in leaves.
The numbers of mesophyll cell layers in leaves, and
consequently leaf thickness, may vary. More layersincrease
the total photosynthetic capacity of the leaf, but upper
layers will shade lower layers, preventing them from get-
ting sufficient light for photosynthesis, especially under
low-light conditions. Because leaf transpirational lossesare
proportional to surface area, the number of mesophyll cell
layers also affects the ratio of photosynthesis to transpi-
ration (or water use efficiency), thus creating a link be-
tween trade-offs related to light and water. The mesophyll
cells of the leaves may contain more chloroplasts to in-
crease their photosynthetic capacity, but this is also as-
sociated with a greater nitrogen investment and greater
maintenance respiration rate. Thus, there are also a num-
ber of trade-offs of the nitrogen and light economies with
the carbon and water economy of the leaf.
The thickness of leaf cell walls relative to cell diameter
limits the maximum pressure that the cell wall can resist
without rupturing. Increasing cell wall thickness or de-
creasing cell diameter can increase themaximumallowable
osmoticpotential of leaf cells, whichiscriticaltotheplant’s
ability to draw in water at low soil moisture. However,
thicker cell walls not only have a greater construction cost
but also increase shading of mesophyll cells. Similarly, in-
creasing leaf cell diameter decreases internal shading but
also decreases the surface-area-to-volume ratio of the cell
and thus reduces the internal conductance to CO2diffu-
sion and thereby reduces photosynthetic rate. Thus, there
are trade-offs between the ability to survive drought and
the resource economies of the leaf.
Leaves may contain varying amounts of sclerenchyma.
Sclerenchyma increases the strength of a leaf. Similarly,
the thickness of mesophyll cell walls relative to mesophyll
cell diameter increases the strength of the leaf at the cost
of increased carbon investment. In general, the longevity
of leaves is a function of their strength, presumably to
resist mechanical damage by herbivores or wear and tear
from wind. Trees also can invest in petioles, the rachis of
compound leaves or twigs, to display their leaves. A de-
ciduous petiole or rachis is less expensive to construct,
whereas twigs may be reused to support new leaves after
the old leaves senesce. There is a general trade-off between
tissue construction cost and longevity.
The complex system of trade-offs for leaves is func-
tionally linked with the trade-offs among the traits of the
support structures such as the stem, branches, and thick
roots. These support structures supply the leaves with the
resources they need in exchange for carbon fixed in the
leaves. For example, a higher hydraulic conductance in-
creases water supply to the leaves, which allows higher
transpiration rates and consequently higher photosyn-
thetic rates. The total stem or branch cross-sectional area
affects this hydraulic conductance but also increases con-
struction cost of the structure. Furthermore, xylem con-
duits that can withstand lower water potentials to survive
droughts need to have thicker cell walls to guard against
implosion in the case of cavitation, and this increases in-
vestment costs per conduit. The relative thickness of the
conduit walls determines wood density and thereby wood
strength. There is a direct link between the hydraulic and
mechanical traits of the wood. In particular, stems can be
58 The American Naturalist
extended in height and crowns can be extended in width
to intercept more light, but as stems and branches become
longer, their construction and maintenance costs also in-
crease, as does the load that needs to be supported me-
chanically by the stem. To support this increasing me-
chanical load, the stems and branches must increase in
cross-sectional area, further increasing construction and
maintenance costs. Thus, in stems, there are three-way
trade-offs among the economies of light, carbon, and wa-
ter, as well as interactions with leaf traits.
As in leaves, in fine roots, there are trade-offs among
resource uptake, nitrogen investment, and maintenance
costs. For instance, fine roots can haveagreaterinvestment
in metabolic activity associated with higher nitrogen con-
tent and higher nutrient uptake rates, but this increases
their maintenance respiration and construction costs. Fur-
thermore, fine roots can be distributed primarily in the
upper soil layers, where nutrients are most abundant but
where evaporation is also greater, or in deeper soil layers,
is typically greater, but then the length of coarse roots
must increase. Thus, there is a three-way trade-off among
nutrient uptake, water uptake, and root constructioncosts.
These diverse trade-offs among traits within an organ
type also interact with each other indirectly at the level of
the whole seedling through the overall carbon, nitrogen,
and water balances. Because these resource balances de-
termine the seedling’s growth and survival, it is clear that
the relationship between any one individual trait and fit-
ness is very complex, with numerous direct as well as
indirect effects. The inclusion of these indirect effects is
an important advantage of the multi-trade-off model ap-
proach over single trade-off models.
Given the complexity of the interactions and trade-offs
among traits in the TAD model, we required a powerful
numerical technique to find optimal solutions. We wrote
a GA for this purpose (Marks and Lechowicz, forthcom-
ing) on the basis of the recommendations for mathemat-
ical efficiency given by Goldberg (1989). GAs are numer-
ical optimization techniques initially developed in analogy
to biological evolution (Goldberg 1989). They are not an
accurate representation of actual biological evolution, al-
though qualitatively they may display many of the same
behaviors and constraints on evolvability (Wagner and Al-
tenberg 1996). For example, our algorithm considers a
population of seedling designs evolving via mutation, re-
combination, and selection, but population size and mu-
tation rate are set for mathematical efficiency, not for bi-
ological realism. Our GA is only a powerful numerical
technique to find possible optima within the potentialtrait
combinations in the TAD model.
To test our alternative design hypothesis, we ran the GA
100 times to search for optimal trait combinations, each
time starting with a different random trait combination.
Our optimality criterion required first that a seedling be
mechanically sound and able to survive in the test envi-
ronment and then that seedling growth over a simulated
period be maximized (Marks and Lechowicz, forthcom-
ing). Each of the 100 replicate runs involved 2,000 cycles
generating variants and selecting among them to insure
that the algorithm had stabilized on an optimum com-
bination of tree seedling trait values, which we will refer
to as a design.
The TAD model can explore alternative designs in a
wide variety of test environments, but all the simulations
reported here were run under the same environmental
conditions. Seedling growth was simulated for 250 days,
with a loam soil, humid air, and high light, under a warm
climate. Soil water was replenished to 95% of field capacity
over the entire soil profile every 30 days. The nitrate input
to the soil was at a rate of 10 g N/m2/yr, typical of a fertile
forest site (Larcher 2003), and the atmospheric CO2con-
centration was 355 ppm, a representativemodernambient.
The simulated environment did not include competing
neighbors, and thus the results correspond to an open-
grown tree seedling. The choice of these relatively benign
environmental conditions for the simulations may seem
arbitrary, but subsequent simulations under a variety of
different conditions includingmorestressfulenvironments
led to the same broad conclusions, even when neighbor-
neighbor competition was included (C. O. Marks, un-
Each of the 100 simulations found a different optimal
combination of trait values representing a viable, alter-
native tree seedling design in the test environment. We
computed the fitness of a design as the simulated growth
and survival of that variant design. If a seedling of a par-
ticular design survived and did not fail mechanically dur-
ing the 250-day simulation, then its fitness is its final dry
mass; otherwise, its fitness is 0. Because the simulation
time was always 250 days, final dry mass is directly cor-
related to average growth rate, a widely accepted measure
of tree seedling fitness. A rank order diagram of fitness
for the 100 designs (fig. 1) indicates that not all the designs
capable of surviving in the test environment have the same
fitness. In other words, the optima found by the GA are
Alternative Functional Designs59
Figure 1: Fitness (seedling dry mass after a 250-day growing period) for the 100 seedling designs where each design representsanoptimalcombination
of trait values found by the genetic algorithm in a simulated growing season. The horizontal line indicates where fitness is 20% lower than the
fitness of the best design. Of the 100 designs, 37 are above this line. We focus on these top 37 designs because they could be considered both well
adapted and approximately equivalent in fitness.
local optima within the trait space, not necessarily global
optima. Despite this range in fitness, a substantial set of
these locally optimal designs converges to a similarly high
level of fitness in the test environment. For example, 37
of the 100 designs had a fitness that was within 20% of
the maximum(fig. 1). Because ourhypothesisisconcerned
with the evolution of alternative designs that are approx-
imately equally fit in terms of seedling growth and we
consider the designs that were not within the top 20% to
be relatively poorly adapted in this regard, we restrict our
subsequent analyses to only these 37 best designs among
the 100 locally optimal designs.
To determine whetherthediverseseedlingdesignsfound
by the GA are indeed representative of truly alternative
designs rather than slightly different versions of the same
general design, we examined the variation in their trait
values. We used a cluster tree based on similarity in trait
values to summarize and assess the patterns in overall trait
variation among the top 37 designs (fig. 2). If evolution
in the model were converging on a single optimal design,
one would expect the designs with the highest fitness to
form a group in the cluster tree, but this is not the case
(fig. 2). This point can be made more explicit by plotting
the fitness difference between design types versus the dis-
tance they are apart in the cluster tree (fig. 3). If there
were convergence, the plot would show a positive rela-
tionship, which it clearly does not. These results imply that
quite comparable fitness can be achieved in substantially
different ways. For example, 23 out of the 34 independent
traits differed by a factor of two or more, and only three
traits had converged completely in the five most-fit de-
signs. It is important to remember that all 34 traits con-
tribute to the fitness of a design. Consequently, trait var-
iation is representative of differences among optima, not
60 The American Naturalist
Figure 2: Cluster tree using simple Euclidean distance and single linkage to compare similarity in the 34 independent traits among the optimal tree
seedling designs (SPSS 1998). Each case represents an optimal design, where the design number represents the rank of that design in terms of fitness.
The values of the 34 independent traits were normalized by the maximum value in the range of possible values for that trait to insure that all traits
are weighted equally in the cluster tree. Only the 37 best seedling designs (i.e., fitness within the top 20%) were included in the analysis because
the focus of the project is on alternative designs that are approximately equally well adapted. For example, note that the five highest-ranked designs
are not all located in the same cluster within the tree. If the results were converging on a single optimal design, one would expect the five best
designs to be located in the same cluster. We highlighted these five designs in the tree by using dashed lines.
as a result of drift in selectively neutral traits. The large
variation in trait values from design to design indicates
that not only do the optima represent distinct functional
designs but also these alternative designs are located far
apart in trait space.
Although the model does not need to be a very accurate
description of tree seedlings to answer our primary ques-
tion, the results are more interesting and convincing if it
can be shown that the optimal seedlings are reasonably
realistic. We found several statistically significant bivariate
relationships between dependent leaf traits (table 1) that
compare well with relationships reported in the literature.
In particular, the globally valid relationships among leaf
mass per area, mass-based leaf nitrogen content, and max-
Alternative Functional Designs61
Figure 3: Plot of fitness difference between the designs in figure 2 versus the distance measure of their trait similarity from the cluster tree. If the
designs were converging on a single design type, a significant positive relationship would be expected in this plot. However, there is actually a
statistically significant ( ) negative relationship that explains !1% of the variation.P ! .05
imum photosynthetic rate were reproduced by the TAD
model (Wright et al. 2004). Other realistic relationships
include the positive relationship between photosynthetic
capacity and xylem sap-flow per leaf area or stomatal con-
ductance (Meinzer 2003), the relationship between the leaf
area index (leaf area per ground area) and maximum net
photosynthetic rate (Canham et al. 1994), the negative
relationship between nitrogen use efficiency and water use
efficiency (Field et al. 1983), and the well-established pos-
itive relationship between area-based maximum photo-
synthetic rate and leaf nitrogen per area (Wright et al.
2004). These relationships between dependent traits are
not programmed into the TAD model but rather arise as
the outcome of interactions among the 34 independent
traits that are under selection. The r2values for some of
these bivariate relationships are relatively low, but this is
also the case for field measurements. This realism in re-
lationships among the dependent traits is reassuring.
Beyond the concern that trait relationships in the model
results should be realistic qualitatively, it is desirable that
the absolute values of traits also be reasonable. We there-
fore compare the range in values for dependent traits with
published data (table A1 in the online edition of the Amer-
ican Naturalist). The values for dependent traits are not
programmed into the model but arise in the optimal res-
olution of trade-offs among the independent traits. The
values for 16 of 19 dependent traits fall well within their
natural ranges; the other three traits are only slightly out-
side their natural range (table A1). In the case of wood
density, and both minimum leaf and minimum xylem wa-
ter potentials, the evolved values were a bit too low and
leaf nitrogen content high because moisture conditions in
the simulations were higher than normal in most envi-
ronments. The traits of the diverse seedling designs iden-
tified in the simulations are consistent with observations
in nature, a conclusion that is further supported in a more
detailed presentation of the TAD model (Marks and
Alternative Functional Designs
Our hypothesis of alternative functional designs as a con-
sequence of multiple interacting trade-offs was supported
by the TAD simulations. The 100 program runs stabilized
on 100 different optima with wide variation in trait values,
suggesting that there may in fact be thousands of quan-
titatively variant optima within this model framework.
This abundance of optimal designs implies that different
lineages evolving in identical environments are highly un-
likely to evolve the same combination of trait values. For
example, when comparing only the five designs with the
highest fitness among the 100 optimal design types, two-
thirds of the 34 independent traits varied by a factor of
two or more. This large variation indicates that these op-
tima represent alternative functional designs and are not
merely slightly different versions of the same design, a
conclusion also supported by the separation of optimal
designs in the cluster analysis (figs. 2, 3).
62 The American Naturalist
Table 1: Linear regression relationships between dependent leaf trait values for the 37 best designs selected
in our test environment
Leaf blade nitrogen concentration per mass Maximum net photosynthetic rate per mass
Leaf mass per leaf areaLeaf blade nitrogen concentration per mass
Leaf mass per leaf area Maximum net photosynthetic rate per mass
Maximum net photosynthetic rate per areaTotal xylem sap flow per leaf area
Maximum net photosynthetic rate per areaSeedling leaf area index
Leaf nitrogen use efficiencyLeaf water use efficiency
Maximum net photosynthetic rate per area Leaf blade nitrogen concentration per area
Maximum net photosynthetic rate per area Maximum realized stomatal conductance
Dependent variableAdjusted r2
Note: Relationships included in the table had
with those observed in field data from natural tree seedlings.
. The r2values for these untransformed relationships are comparableP ! .01
A number of models of plant form and function have
found alternative designs, but the alternative designs were
generally interpreted as a result of differences in selection
regime associated with frequency dependence or environ-
mental differences (Farnsworth and Niklas 1995; Niklas
1997b; Schwinning and Ehleringer 2001; Warren and Top-
ping 2001; Mustard et al. 2003). For example, in an elegant
model, Niklas (1994, 1997a, 1999) showed that trees can
have alternative architectures depending on the relative
importance of fitness of mechanical stability, seed disper-
sal, and light interception. In contrast, reviews on adap-
tation have mentioned the idea that alternative designs
might be fit in the same environment for a long time(Bock
1959, 1976, 1980; Lewontin 1978; Gould and Lewontin
1979; Mayr 1983; Gould 1989; Parker and Maynard Smith
1990; Ko ¨rner 1991; Losos and Miles 1994; Niklas 1997b;
Gutschick 1999). Most of these reviews discuss alternative
designs conceptually and give few if any examples, which
can give readers the impression that alternative designs are
a rare occurrence in nature—only an interesting exception
in a general trend toward convergence to a single optimal
design in a given environment. Our results demonstrate
that there is such a large potential for alternative designs
in the same environment that different lineages almost
inevitably will evolve toward different optimal designs.
Furthermore, alternative designs dramatically affect most
functional traits, suggesting that aside from environmental
heterogeneity, alternative designs are a major, if not the
main, cause of functional diversity in nature.
Alternative designs also are recognized in a number of
models of abiotic complex systems. For example, in ap-
plications of GAs to engineering design problems,multiple
alternative designs typically are found, including ones that
are very similar to successful designs previously developed
independently by human designers (Bentley 1999; Koza et
al. 1999). Similarly, in the evolution of simple artificial life
forms, multiple designs usually evolve (Kauffman 1993;
Sims 1999; Yedid and Bell 2002; Chow et al. 2004). What
is unique about our results is that they were obtained with
a biologically realistic model of a higher organism inwhich
design outcomes can be directly compared with empirical
data (Marks and Lechowicz, forthcoming).
It is likely that there is a general mathematical basis to
the emergence of alternative designs in any complex sys-
tem involving interactions among traits. Consider a very
simple model involving only two individual traits (A and
B) and a composite trait (C) that is a function of traits A
and B. For example, assume
C at some constant intermediate value maximized fitness
(i.e., stabilizing selection on C), then there obviously are
multiple alternative combinations for the values of A and
B that yield the optimal value for C—any increase in the
value of A could be compensated for by a decrease in the
value for B. Consequently, there would be an infinitenum-
ber of optimal combinations for the values of traits A and
B. Extending this simple example to more trade-offs
among a greater number of traits, it is clear there is no
theoretical limit to the number of potential alternative
designs that might emerge in a complex system.
A detailed biological example on these same lines re-
illustration of the alternative design concept (Alfaro et al.
2004, 2005). Labrid fish jaws consist of a four-bar mech-
anism. The shape of the four-bar affects the maximal ki-
nematic transmission coefficient (or max KT) in a redun-
dant fashion, and max KT is significantly correlated with
labrid fish ecology. Consequently, there are multiple com-
Any complex system of trait interactions comprising a
holistic model of organism function contains numerous
examples of similar measures of performance that depend
on multiple independent traits. For example, in our TAD
model, seedling water use depends on root traits for water
uptake, xylem traits for water conduction, and stomatal
control traits for regulation. Similarly, the nitrogen and
. If keeping traitC p A ? B
Alternative Functional Designs 63
light economy are affected by several traits. The most dra-
matic example is the carbon economy, which is affected
by every structure within the plant because all structures
have a construction and maintenance cost. Furthermore,
survival and growth and thus plant fitness depend on the
interaction among the water, nitrogen, light, and carbon
use strategies. Consequently, there is a large potential for
alternative functional designs in tree seedlings, and a sim-
ilarly high potential canbe expectedinall organismswhose
performance depends on the complex interaction among
Approximate Fitness Equivalence of Alternative Designs
Many of the optimal designs found by the GA in our test
environment had a similar high fitness (fig. 1); more than
one-third of the optimal designs reached a fitness within
the top 20% of the range. These alternative designs exploit
available resources in different ways but to the same net
effect in terms of growth. This provides a potential ex-
planation for the frequent observation of saplings growing
equally well at the same site despite large trait differences.
In the forestry literature, there are many records of co-
occurring tree species capable of similar growth rates de-
spite substantial trait differences (Goulden 1996; Oliver
and Larson 1996; Becker et al. 1999; Valladares et al. 2002;
Cavender-Bares et al. 2004b). We expect that future re-
search will reveal many examples of equicompetent alter-
native designs, particularly in the more diverse forests of
Although it is not feasible to do laboratory experiments
with tree evolution to test the alternative design hypothesis
directly, the results of experimental evolution studies with
bacteria are remarkably similar to the results of our mod-
eling study with tree seedlings (Travisano et al. 1995; Ko-
rona 1996; Nakatsu et al. 1998; Velicer and Lenski 1999;
MacLean and Bell 2003). For example, Korona (1996)
found that six replicates of a bacterial culture derived from
the same clone evolvingunderidenticalconditionsreached
the same fitness level once the cultures stabilized. The
fitness increase was proportional to an increase in growth
rate, but in eachcasedifferenttraitcombinationsproduced
this increase. MacLean and Bell (2003) found that in an
adaptive radiation of a bacterium to different growth me-
dia, there was no correlation between medium and genetic
similarity of the optimal types, implying that the necessary
adaptive change was achieved in many different ways. The
similarity of results for organisms as different as bacteria
and trees suggests that multiple optima of approximately
equal growth rate likely also occur in other groups of
organisms that display great trait diversity within habitats.
The large potential for the evolution of diverse alternative
designs has some interesting ecological implications. In
particular, we found that species representing alternative
designs might nonetheless differ by only a few percent in
fitness (i.e., growth rate). Although such small differences
in fitness should promote species coexistence, indefinite
coexistence would still require a stabilizing mechanism
such as density- or frequency-dependent mortality (Ches-
son 2000). However, when fitness differences are small, as
in the best designs in our results, even weak stabilizing
mechanisms are sufficient to insure coexistence (Chesson
2000), and field ecologists have elucidated an abundance
of such weak stabilizing mechanisms. In this sense, the
existence of alternative designs of approximately equiva-
lent fitness not only supports the evolution of diversity in
functional plant traits but in conjunction with weak sta-
bilizing mechanisms also augments species diversitywithin
relatively uniform habitats.
The model results we presented provide an adaptive ex-
planation for the frequent observation of tree seedlings
growing equally well at the same site despite large trait
differences among species. Such an explanation is timely
because the widespread occurrence of large niche overlap
is receiving more and more attention in field studies(Rick-
lefs 2000; Pitman et al. 2001; Valencia et al. 2004) and in
theoretical models (Bell 2000; Hubbell 2001, 2005). We
suspect that niche overlap is as important in community
assembly as are niche differences and thatthislargeoverlap
in niches is at least in part a consequence of the evolution
of alternative functional designs.
According to conventional wisdom, trait diversity
evolves only in response to environmental heterogeneity
as predicted by the single trade-off optimization models
(Levin and Muller-Landau 2000), but our holistic model
demonstrates that even in the absence of heterogeneity,
separate lineages evolve large trait differences. Therefore,
alternative functional designs provide a convincing expla-
nation for the large within-site trait variation observed in
forests all over the world.
We would like to thank D. Ackerly, G. Bell, J. Cavender-
Bares, A. Gonzalez, D. Schluter, D. Tilman, and two anon-
ymous reviewers for commenting on an earlier version of
the manuscript. We also thank K. Arii, M. Futer, T.
Linkosalo, and R. Roy for their support. We are grateful
to the Natural Sciences and Engineering Research Council
(NSERC), McGill University, and Groupe de Recherche
64 The American Naturalist
en E´cologie Forestie `re Interuniversitaire for generously
providing scholarships for C.O.M. The research was
funded by an NSERC grant to M.J.L.
Aerts, R., and F. S. Chapin. 2000. The mineral nutrition of wild
plants revisited: a re-evaluation of processes and patterns. Ad-
vances in Ecological Research 30:1–67.
Alfaro, M. E., D. I. Bolnick, and P. C. Wainwright. 2004.Evolutionary
dynamics of complex biomechanical systems: an example using
the four-bar mechanism. Evolution 58:495–503.
———. 2005. Evolutionary consequences of many-to-one mapping
of jaw morphology to mechanics in labrid fishes. American Nat-
Becker, P., M. T. Tyree, and M. Tsuda. 1999. Hydraulic conductances
of angiosperms versus conifers: similar transport sufficiency at the
whole-plant level. Tree Physiology 19:445–452.
Bell, G. 2000. The distribution of abundance in neutralcommunities.
American Naturalist 155:606–617.
Bentley, P. J. 1999. From coffee tables to hospitals: generic evolu-
tionary design. Pages 405–423 in P. J. Bentley, ed. Evolutionary
design by computers. Morgan Kaufmann, San Francisco.
Bock, W. J. 1959. Preadaptation and multiple evolutionarypathways.
———. 1976. Adaptation and the comparative method. Pages 57–
82 in M. K. Hecht, P. C. Goody, and B. M. Hecht, eds. Major
patterns in vertebrate evolution. Plenum, New York.
———. 1980. The definition and recognition of biological adapta-
tion. American Zoologist 20:217–227.
Burns, R. M., and B. H. Honkala. 1990. Silvics of North America.
USDA Forest Service, Washington, DC.
Canham, C. D., A. C. Finzi, S. W. Pacala, and D. H. Burbank. 1994.
Causes and consequences of resource heterogeneity in forests: in-
terspecific variation in light transmission by canopy trees. Cana-
dian Journal of Forest Research 24:337–349.
Caspersen, J. P., and R. K. Kobe. 2001. Interspecific variation in
sapling mortality in relation to growth and soil moisture. Oikos
Cavelier, J. 1990. Tissue water relations in elfin cloud forest tree
species of Serrania de Macuira, Guajira, Colombia. Trees 4:155–
Cavender-Bares, J., K. Kitajima, and F. A. Bazzaz. 2004a. Multiple
trait associations in relation to habitat differentiation among 17
Floridian oak species. Ecological Monographs 74:635–662.
Cavender-Bares, J., D. D. Ackerly, D. A. Baum, and F. A. Bazzaz.
2004b. Phylogenetic overdispersion in Floridian oak communities.
American Naturalist 163:823–843.
Chesson, P. 2000. Mechanisms of maintenance of species diversity.
Annual Review of Ecology and Systematics 31:343–366.
Choong, M. F., P. W. Lucas, J. S. Y. Ong, B. Pereira, H. T. W. Tan,
and I. M. Turner. 1992. Leaf fracture toughness and sclerophylly:
their correlations and ecological implications. New Phytologist
Chow, S. S., C. O. Wilke, C. Ofria, R. E. Lenski, and C. Adami. 2004.
Adaptive radiationfrom resourcecompetitionindigitalorganisms.
Comas, L. H., T. J. Bouma, and D. M. Eissenstat. 2002. Linking root
traits to potential growth rate in six temperate tree species.
Oecologia (Berlin) 132:34–43.
Ehleringer, J. R., E.-D. Schulze, H. Ziegler, O. L. Lange, G. D. Far-
quhar, and I. R. Cowan. 1985. Xylem-tapping mistletoes: water or
nutrient parasites? Science 227:1479–1481.
Ellis, G. F. R. 2005. Physics, complexity and causality. Nature 435:
Espeleta, J. F., and L. A. Donovan. 2002. Fine root demography and
morphology in response to soil resources availability among xeric
and mesic sandhill tree species. Functional Ecology 16:113–121.
Farnsworth, K. D., and K. J. Niklas. 1995. Theories of optimization,
form and function in branching architecture in plants. Functional
Field, C., J. Merino, and H. A. Mooney. 1983. Compromises between
water-use efficiency and nitrogen-use efficiency in five species of
California evergreens. Oecologia (Berlin) 60:384–389.
Forest Products Laboratory. 1999. Wood handbook: wood as an en-
gineering material. USDA Forest Service General Technical Report
FPL-GTR-113, Madison, WI.
Fralish, J. S., S. M. Jones, R. K. O’Dell, and J. L. Chambers. 1978.
The effect of soil moisture on site productivity and forest com-
position in the Shawnee hills of southern Illinois. Pages 263–285
in W. E. Balmer, ed. Proceedings of the Soil Site Productivity
Symposium. USDA Forest Service, Washington, DC.
Goldberg, D. E. 1989. Genetic algorithms in search, optimization,
and machine learning. Addison-Wesley, Reading, MA.
Gould, S. J. 1989. Wonderful life: the Burgess shale and the nature
of history. Norton, New York.
Gould, S. J., and R. C. Lewontin. 1979. The spandrels of San Marco
and the Panglossian paradigm: a critique of the adaptationist pro-
gramme. Proceedings of the Royal Society of London B 205:581.
Goulden, M. L. 1996. Carbon assimilation and water-use efficiency
by neighboring Mediterranean-climate oaks that differ in water
access. Tree Physiology 16:417–424.
Groom, P. K. 2004. Rooting depth and plant water relations explain
species distribution patterns within a sandplain landscape. Func-
tional Plant Biology 31:423–428.
Gutschick, V. P. 1999. Biotic and abiotic consequences of differences
in leaf structure. New Phytologist 143:3–18.
Hacke, U. G., J. S. Sperry, W. T. Pockman, S. D. Davis, and K. A.
McCulloh. 2001. Trends in wood density and structure are linked
to prevention of xylem implosion by negative pressure. Oecologia
Harcombe, P. A. 1987. Tree life tables: simple birth, growth, and
death data encapsulate life histories and ecological role.BioScience
Holbrook, N. M., and F. E. Putz. 1996. From epiphyte to tree: dif-
ferences in leaf structure and leaf water relations associated with
the transition in growth form in eight species of hemiepiphytes.
Plant Cell and Environment 19:631–642.
Hubbell, S. P. 2001. The unified neutral theory of biodiversity and
biogeography. Princeton University Press, Princeton, NJ.
———. 2005. Neutral theory in community ecology and the hy-
pothesis of functionalequivalence.FunctionalEcology19:166–172.
Iverson, L. R., A. M. Prasad, B. J. Hale, and E. Kennedy Sutherland.
1999. Atlas of current and potential future distributions of com-
mon trees of the eastern United States. USDA Forest Service,
Northeastern Research Station, Radnor, PA.
Iwasa, Y., D. Cohen, and J. A. Leon. 1984. Tree height and crown
shape, as results of competitive games. Journal of Theoretical Bi-
Alternative Functional Designs65
Kauffman, S. A. 1993. The origins of order: self-organization and
selection in evolution. Oxford University Press, New York.
Kobe, R. 1997. Carbohydrate allocation to storage as a basis of in-
terspecific variation in sapling survivorship and growth. Oikos 80:
Ko ¨rner, C. 1991. Some often overlooked plant characteristics as de-
terminants of plant growth: a reconsideration. Functional Ecology
Korona, R. 1996. Genetic divergence and fitness convergence under
uniform selection in experimental populations of bacteria. Ge-
Koza, J., F. H. Bennett, D. Andre, and M. A. Keane. 1999. The design
of analogue circuits by means of genetic programming. Pages365–
385 in P. J. Bentley, ed. Evolutionary design by computers.Morgan
Kaufmann, San Francisco.
Lambers, H., F. S. Chapin, and T. L. Pons. 1998. Plant physiological
ecology. Springer, New York.
Larcher, W. 2003. Physiological plant ecology. Springer, Berlin.
Levin, S. A., and H. C. Muller-Landau. 2000. The emergence of
diversity in plant communities. Comptes Rendus de l’Acade ´mie
des Sciences III 323:129–139.
Lewontin, R. C. 1978. Adaptation. Scientific American 239:213–230.
Losos, J. B., and D. B. Miles. 1994. Adaptation, constraint, and the
comparative method: phylogenetic issues and methods. Pages 60–
98 in P. Wainwright and S. Reilly, eds. Ecological morphology:
integrative organismal biology. University of Chicago Press,
MacLean, R. C., and G. Bell. 2003. Divergent evolution during an
experimental adaptive radiation. Proceedings of the Royal Society
of London B 270:1645–1650.
Maherali, H., W. T. Pockman, and R. B. Jackson. 2004. Adaptive
variation in the vulnerability of woody plants to xylem cavitation.
Ma ¨kela ¨, A., T. J. Givnish, F. Berninger, T. N. Buckley, G. D. Farquhar,
and P. Hari. 2002. Challenges and opportunities of the optimality
approach in plant ecology. Silva Fennica 36:605–614.
Marks, C. O., and M. J. Lechowicz. Forthcoming. A holistic tree
seedling model for the investigation of functional trait diversity.
Mayr, E. 1983. How to carry out the adaptionist program? American
Meinzer, F. C. 2003. Functional convergence in plant responses to
the environment. Oecologia (Berlin) 134:1–11.
Mustard, M. J., D. B. Standing, M. J. Aitkenhead, D. Robinson, and
A. J. S. McDonald. 2003. The emergence of primary strategies in
evolving virtual-plant populations. Evolutionary EcologyResearch
Nakatsu, C. H., R. Korona, R. E. Lenski, F. J. de Bruijn, T. L. Marsh,
and L. J. Forney. 1998. Parallel and divergent genotypic evolution
in experimental populations of Ralstonia sp. Journal of Bacteri-
Niinemets, U¨. 2001. Global-scale climatic controls of leaf dry mass
per area, density, and thickness in trees and shrubs. Ecology 82:
Niklas, K. J. 1994. Morphological evolution through complex do-
mains of fitness. Proceedings of the National Academy of Sciences
of the USA 91:6772–6779.
———. 1997a. Adaptive walks through fitness landscapes for early
vascular land plants. American Journal of Botany 84:16–25.
———. 1997b. The evolutionary biology of plants. University of
Chicago Press, Chicago.
———. 1999. Evolutionarywalksthroughalandplantmorphospace.
Journal of Experimental Botany 50:39–52.
Nobel, P. S. 1991. Physicochemical and environmental plant physi-
ology. Academic Press, San Diego, CA.
Oliver, C. D., and B. C. Larson. 1996. Forest stand dynamics. Wiley,
Pacala, S. W., C. D. Canham, J. A. Silander Jr., and R. K. Kobe. 1994.
Sapling growth as a function of resources in a north temperate
forest. Canadian Journal of Forest Research 24:2172–2183.
Parker, G. A., and J. Maynard Smith. 1990. Optimality theory in
evolutionary biology. Nature 348:27–33.
Pitman, N. C. A., J. W. Terborgh, M. R. Silman, P. Nu ´n ˜ez V., D. A.
Neill, C. E. Cero ´n, W. A. Palacios, and M. Aulestia. 2001. Dom-
inance and distribution of tree species in upper Amazonian terra
firme forests. Ecology 82:2101–2117.
Prior, L. D., D. Eamus, and D. M. J. S. Bowman. 2003. Leaf attributes
in the seasonally dry tropics: a comparison of four habitats in
northern Australia. Functional Ecology 17:504–515.
Proulx, S. R., D. E. L. Promislow, and P. C. Phillips. 2005. Network
thinking in ecology and evolution. Trends in Ecology & Evolution
Reich, P. B., M. G. Walters, D. Tjoelker, D. Vanderklein, and C.
Buschena. 1998. Photosynthesis and respiration rates depend on
leaf and root morphology and nitrogen concentration in nine bo-
real tree species differing in relative growth rate. Functional Ecol-
Reich, P. B., D. S. Ellsworth, M. B. Walters, J. M. Vose, C. Gresham,
J. C. Volin, and W. D. Bowman. 1999. Generality of leaf trait
relationships: a test across six biomes. Ecology 80:1955–1969.
Ricklefs, R. E. 2000. Rarity and diversity in Amazonian forest trees.
Trends in Ecology & Evolution 15:83–84.
Schluter, D. 2000. Ecological character displacement in adaptive ra-
diation. American Naturalist 156(suppl.):S4–S16.
Schwinning, S., and J. R. Ehleringer. 2001. Water use trade-offs and
optimal adaptations to pulse-driven arid ecosystems. Journal of
Sims, K. 1999. Evolvingthree-dimensionalmorphologyandbehavior.
Pages 297–321 in P. J. Bentley, ed. Evolutionary design by com-
puters. Morgan Kaufmann, San Francisco.
SPSS. 1998. SYSTAT. Version 9. SPSS, Chicago.
Sutherland, W. J. 2005. The best solution. Nature 435:569.
Travisano, M., J. A. Mongold, A. F. Bennett, and R. E. Lenski. 1995.
Experimental tests of the roles of adaptation, chance, and history
in evolution. Science 267:87–90.
Valencia, R., R. B. Foster, G. Villa, R. Condit, J.-C. Svenning, C.
Herna ´ndez, K. Romoleroux, E. Losos, E. Maga ˚rd, and H. Balslev.
2004. Tree species distributions and local habitat variation in the
Amazon: large forest plot in eastern Ecuador. Journal of Ecology
Valladares, F., J. B. Skillman, and R. W. Pearcy. 2002. Convergence
in light capture efficiencies among tropicalforestunderstoryplants
pensation. American Journal of Botany 89:1275–1284.
Van Valen, L. 1975. Life, death, and energy of a tree. Biotropica 7:
Velicer, G. J., and R. E. Lenski. 1999. Evolutionary trade-offs under
conditions of resource abundance and scarcity: experiments with
bacteria. Ecology 80:1168–1179.
66 The American Naturalist Download full-text
Wagner, G. P., and L. Altenberg. 1996. Complex adaptations and the
evolution of evolvability. Evolution 50:967–976.
Warren, J., and C. Topping. 2001. Trait evolution in an individual-
based model of herbaceous vegetation. Evolutionary Ecology 15:
Westoby, M., D. S. Falster, A. T. Moles, P. A. Vesk, and I. J. Wright.
2002. Plant ecological strategies: some leading dimensions of var-
iation between species. Annual Review of Ecology and Systematics
Whittaker, R. H. 1956. Vegetation of the Great Smoky Mountains.
Ecological Monographs 26:1–80.
Wright, I. J., P. B. Reich, M. Westoby, D. D. Ackerly, Z. Baruch, F.
Bongers, J. Cavender-Bares, et al. 2004. The worldwide leaf eco-
nomics spectrum. Nature 428:821–827.
Yedid, G., and G. Bell. 2002. Macroevolution simulated with auton-
omously replicating computer programs. Nature 420:810–812.
Associate Editor: Mark Westoby
Editor: Jonathan B. Losos