Review TRENDS in Ecology and Evolution
Why evolutionary biologists should be
demographers
C. Jessica E. Metcalf1,2 and Samuel Pavard1
1
Max Planck Institute for Demographic Research, Konrad-Zuse Str. 1, D-18057, Rostock, Germany
2
Department of Biology, 139 Biological Sciences Dr, Duke University, Durham, NC 27707, USA
Evolution is driven by the propagation of genes, traits
and individuals within and between populations. This
propagation depends on the survival, fertility and dis-
persal of individuals at each age or stage during their life
history, as well as on population growth and (st)age
structure. Demography is therefore central to under-
standing evolution. Recent demographic research pro-
vides new perspectives on fitness, the spread of
mutations within populations and the establishment
of life histories in a phylogenetic context. New chal-
lenges resulting from individual heterogeneity, and
instances where survival and reproduction are linked
across generations are being recognized. Evolutionary
demography is a field of exciting developments through
both methodological and empirical advances. Here, we
review these developments and outline two emergent
research questions.
Introduction
Demography, the study of survival, fertility and population
dynamics, is a crucial tool for evolutionary biologists. In
particular, survival and fertility at each age or life-history
stage determine offspring production, which defines fit-
ness. However, the situation is complex: fitness can be
estimated for a single individual, a subpopulation of indi-
viduals sharing a genotype or phenotype, or an entire
population, in constant or fluctuating environments and
in density-dependent or independent contexts (Table 1).
This diversity makes the appropriate fitness measure a
difficult topic. It also complicates the ultimate goal of
evolutionary ecology, that is, to understand the establish-
ment of a given life history and life-history diversity across
species. Although more rarely considered, complications
also emerge, because spatial structure (dispersal) and
genetic drift might also affect life-history evolution. Recent
work also indicates that the heterogeneity of survival and
fertility among individuals and across generations can
profoundly modify estimates of individual fitness.
This brief overview makes understanding life-history
evolution seem dauntingly complex. However, the differ-
ent sources of complexity are theoretically and empirically
exciting and shed new light on where and why demogra-
phy matters in evolution. The recent availability of large
long-term data sets, enabling estimation of survival and
fertility across (st)ages in several species [i.e. (st)age
Corresponding author: Metcalf, C.J.E. (cjm29@duke.edu).
Available online 13 December 2006.
www.sciencedirect.com 0169-5347/$ see front matter 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.tree.2006.12.001
Vol.22 No.
trajectories of survival and fertility], has led to a flurry
of demographic modelling and empirical exploration of
several evolutionary theories, addressing, for example,
how variation in stage trajectories of survival and fertility
across individuals and between years drives selection on
timing of reproduction in monocarpic plants [1]; when
variation in stage trajectories between years selects for
buffering of individual variation in stage trajectories in a
perennial plant [2]; when covariation between fitness
components affects selection on age trajectories in red
Glossary
Adaptive dynamics: extends invasibility to consider the long-term outcome of
a selective process [the evolutionarily stable strategy (EES)] and the process
involved in attaining this outcome (mutation and iterative invasion).
Effective population size (Ne): the size of an ideal (stable, random mating)
population that results in the same degree of genetic drift or inbreeding as
observed in the actual population.
Ergodic: the dynamic of a population when, after an interval of time, the
population aysmptotically reaches a stable (st)age distribution and a
correspondingly constant population growth rate, which is independent of
the initial (st)age distribution.
Hierarchical Bayes: a set of statistical tools using relationships between
conditional probabilities defined by Bayes Rule to break complex statistical
distributions into sets of interdependent distributions for which parameters
can be obtained iteratively using Markov Chain Monte Carlo approaches. It is
often used to model data resulting from complex underlying processes, such
as interdependence in demographic rates.
Integral projection models: analogous to matrix population models, but can be
based on a continuous rather than discrete population structure (such as size
rather than life-history stage). Population structure is expressed as a density
function, so that inclusion of variance within or across individuals is
straightforward and can be directly related to standard statistical techniques.
All population measures available from matrix population models can also be
obtained from integral projection models.
Invasibility analysis: resembles optimization, except that instead of comparing
fitness measures, a population context is modelled. Repeated invasion of a
chosen resident strategy (which sets some component of the environment for
the invader) is used to identify the strategy that cannot be invaded by any
other, or the ESS. Invasion is considered successful for strategies that can
invade when rare. Successful invasion is identified using invasion exponents
(Table 1).
Matrix population models: can be stage or age-based and enable calculation of
R0, l, ls, or invasion exponents (Table 1), as well as their stage- or age-specific
sensitivities or elasticities; or sensitivity or elasticity to underlying parameters
[56,7]. Other population characteristics, such as the stable (st)age structure or
reproductive value, are available.
Optimization: involves defining a fitness measure and then locating the
strategy or trait that maximizes it given a set of constraints.
Perturbation analysis: involves calculating changes in fitness corresponding to
a relative (i.e. elasticity) or absolute (i.e. sensitivity) change in a life-cycle
component. This enables quantification of the relative or absolute contribu-
tions of a life-cycle component or life-cycle path to population growth rate and
thereby fitness.
Transient: term used to define population dynamics experienced by
populations when the (st)age distribution fluctuates unstably. Transient
fluctuations in (st)age structure can persist indefinitely in non-ergodic systems.
The appropriate fitness measure for transient phases of population dynamics
is still unclear.
206 Review TRENDS in Ecology and Evolution Vol.22 No.4

deer [3]; and how individual contributions to population
growth rate from survival and fecundity can be estimated
in Soay sheep [4].
Such theoretical and empirical developments make it
timely to review how new demographic research clarifies
evolutionary questions. In particular, we address here how
changes in (st)age trajectories of survival and fertility
affect (i) fitness; (ii) the spread of a mutation within a
population; and (iii) the dynamic interplay between age
trajectories and selection pressures that lead to the estab-
lishment of a particular life history over many generations.
We then choose to discuss two major challenges in evolu-
tionary biology where a demographic perspective is essen-
tial. First, (st)age trajectories vary across individuals and
environments, which hampers the accurate inference of
(st)age trajectory parameters and can alter evolutionary
outcomes. Second, survival and reproduction might be
linked across generations, which has major implications
for the evolution of parental care. Both cases call for
progress in the development of more complex models.

Table 1. Different fitness measures, both for direct comparison and in a population context (invasion)
Fitness measure Formula Description Where appropriate Refs
I. Stable population theory (based on ergodic assumptions)
Constant environments
X
Comparison Net R0 l mx a,b
x x
Average number of offspring born to Constant environments [43,56]
reproductive X individuals during their lifetime; or
T xl mx a,b Non-overlapping
rate, R0, and x x factor by which the size of the generations
generation population is multiplied after one Rate insensitive c
time, T generation of length T
Finite rate of Root of the EulerLotka equation: Finite increase in population size per Constant environments [43,56]
X
increase, l lx l x mx 1 unit of absolute time Overlapping or non-
overlapping generations
Rate sensitive c
Invasion Invasion Similar to l but where lx or mx Finite rate of increase of an invader As for l d [56,59,
exponent, W might be functions of population into an environment set by a resident 60]
size N or population size in a strategy affecting the invader through When density or frequency
(st)age class density or frequency dependence dependence are operating e
Fluctuating environments
Comparison Stochastic 1 Stochastic or periodic equivalent of l, Fluctuating environments [2,9,
loglS lim E logN  f
growth rate, lS t !1 t obtained numerically by iteration of 45
matrix multiplication, where matrices 46,56]
for every time-step t are defined
by environmental conditions
Invasion Invasion 1 Rate of invasion of a rare invader Fluctuating environments [56]
# lim E logN i  f
exponent, W t !1 t into an environment set by both
fluctuations in the environment, When density or frequency
and the dynamics of a resident dependence are operating
strategy (acting through frequency
or density dependence)g
II. Unstable population theory (no ergodic assumptions)
Comparison De-lifing N t1  zti h Statistical estimate of the Accounts for non-ergodic [4]
P ti w t 
Nt  1 contribution of an individual population dynamics in
to population growth fluctuating environments
Assumes individuals are
independent (i.e. cannot
include frequency or
density dependence)
Transient dnt 1 dnt  T  dvecA j Captures the response of a chosen Accounts for non-ergodic [7]
A n t I S
sensitivities i duT duT duT fitness measure (population size at t, population dynamics in
cumulative density at t, etc.) to fluctuating environments
transient dynamics Requires, and is dependent
on, a definition of
population structure at t0
Has no single fitness
definition, or synthetic
measure
Invasion Transient As above, but where A also Captures the response of a chosen As above, but where the [7]
sensitivities i depends on some aspect of fitness measure to transient population context is
population size N dynamics in the case of density important
or frequency dependence
Individual measures of fitness (not based on population averages)
X
Individual finite F l lI 1 1 k
x x
Equivalent to l but based on Where individual-level [57,58]
rate of increase, individual life-history parameters covariation is expected to
lI (i.e. lx is replaced by 1 for each bias population-level
age x at which the individual averages
survived, and zero otherwise)l,m Problematic, owing to
biases inherent in making
predictions based on a
sample size of 1 l
www.sciencedirect.com
 Review TRENDS in Ecology and Evolution Vol.22 No.4 207

Table 1 (Continued )
Fitness measure Formula Description Where appropriate Refs
X
Lifetime LRS I F I k
x x
Total number of offspring If the population is stable, [13,61]
reproductive produced by each individualm E(LRSI) = R0
success, LRSI
a
x, age; lx, probability of surviving to x; mx, fertility at x.
b
See also Ref. [56].
c
R0 is rate insensitive, so does not capture the importance in differences in relative timing of life-history events; l is rate sensitive and does.
d
If a density-dependent population has a stable equilibrium, the sensitivity of W to underlying parameters will be identical to the sensitivity of l [59].
e
A key result for invasion analysis in constant environments is that if density dependence acts on offspring establishment, the strategy that maximizes R0 is the EES (i.e. cannot
be invaded); if density dependence acts on survival at all ages, the strategy that maximizes l is the ESS [60].
f
E(y), expected value of y; N, population size; Ni, invader population size.
g
For example, if density dependence acts on offspring establishment, the offspring number produced by the invader at each t will be determined by the environmental
conditions at t, and the offspring production of the resident at t [1].
h
pti, contribution of individual i to pop growth during time step t; wt, population size at t + 1 minus population size at t; jt(i), performance of individual i (number surviving
offspring + 1 over time-step t).
i
No direct fitness measure is available but sensitivities can be calculated.
j
A, matrix of population transitions; u, lower level parameter to which sensitivity is being estimated; Is, identity matrix; vec, operator that stacks matrix columns into a column
vector.
k
Fx, fertility at age x of individual I.
l
See Ref. [58] for more complex Bayesian approaches.
m
See Ref. [13] for a comparison between LRSI and lI and Ref. [61] for comparison with the long-term genetic contribution.

Estimating fitness How can we quantify fitness?

When will a mutation affect individual fitness?
Even if a mutation changes the phenotype of an individual
in a given environment (through action at any level:
molecular, cellular, physiological or behavioral), it will
not alter fitness unless it changes how the individual
survives and reproduces across (st)ages. Furthermore, if
the mutation acts at a specific age, the degree to which it
alters fitness depends on the survival and fertility of the
affected individual at all other ages. For example, if mor-
tality is such that few individuals are alive beyond a
certain age, and the contribution of these individuals to
lifetime reproduction is negligible, mutations acting after
this age are effectively neutral. Consequently, late-acting
deleterious mutations are subject to less selection and
their resulting accumulation is one possible explanation
for the evolution of senescence.
How do researchers explore fitness in natural
populations?
Empirical work often focuses on the relationship between
a trait and a single fitness component, such as mean age at
first reproduction, life expectancy, total fertility rate,
clutch size, offspring survival, length of the juvenile per-
iod, adult survival, viability, and so on (reviewed in Ref.
[5]). However, traits can be negatively or positively corre-
lated with more than one fitness component. For example,
Coulson et al. [3] showed that birth weight in red deer was
affected by selection on several fitness components,
including birth rate and winter survival of male calves.
Using a single fitness component as a proxy for fitness can,
consequently, be problematic. It is particularly advisable
to avoid using the offspring number of breeding indivi-
duals as a fitness measure. Tradeoffs between juvenile
survival and adult fertility are probable, and a trait that
impacts positively on this fitness measure (i.e. adult fer-
tility) might also impact negatively on juvenile survival.
This is referred to by Grafen [6] as the problem of the
invisible fraction; that is, the fraction of individuals that
die before the trait is measured. Accurate measures of
fitness should therefore (where possible) include both
survival and fertility.
www.sciencedirect.com
To compare the fitness of phenotypes or genotypes, a range
of different fitness measures that summarize the full (st)age
trajectories of survival and fertility is available, including
novel techniques allied to recent approaches in demographic
modeling, reviewed in Table 1. These fitness measures
can be based on individual values or population averages,
in constant or fluctuating environments, stable population
contexts or during transient dynamics. Each fitness mea-
sure has advantages and drawbacks. Determining the
appropriate fitness measure is a complex and controversial
topic, but one that is of considerable importance, as different
fitness measures can lead to different predictions. For exam-
ple, the predicted evolutionarily stable flowering size for the
plant Onopordum illyricum is different in a stochastic
environment compared to that in a constant environment
and is closer to the observed value [1]. Until recently, most
fitness measures have been defined within the framework of
stable population theory (Table 1), but it is increasingly
recognized that transient dynamics might be important.
This is an area of active research [4,7,8].
Once a fitness measure has been chosen, the impact of
changes in fitness components (traits) on the fitness mea-
sure can be analyzed using perturbation analysis (see
Glossary). This basic framework has been considerably
extended in recent years: covariation between life-cycle
components can now be incorporated into elasticity ana-
lysis (e.g. Ref. [3]); the calculation of elasticity in changing
environments has recently been elucidated [9]; and even
more recently, methods for the calculation of elasticity
during transient phases have been developed [7]
(Table 1). However, interpretation of perturbation analysis
is still debated [9] and, again, can vary with the context
considered. One key complication that we now address is
the population context itself [59].
Short-term evolutionary outcomes: the spread of genes
within populations
Earlier, we reviewed the list of classic and recently
developed fitness measures (Table 1). However, to
determine whether a genotype will spread to fixation,
coexist, or disappear when in competition with other
208 Review TRENDS in Ecology and Evolution Vol.22 No.4
genotypes requires more than a comparison of fitness
values. Short-term evolution is driven by selection, migra-
tion and drift. The dynamic interplay among the survival
and fertility of different individuals (e.g. through
density dependence) will impact selection; but also has
implications for drift and migration (more rarely
addressed). Given accumulating empirical evidence for
complex interactions among individuals and populations
in the demography of natural populations [10] it is timely
to consider integrating such complexity into an evolution-
ary perspective.
Selection
The population context affects selection because the fitness
of an individual is contingent on what other individuals in
the population are doing, through: (i) density dependence:
if a mutation improves the fertility of a plant, but only
when it is not shaded by its neighbors, the mutation is
unlikely to spread successfully through a high-density
population (e.g. Ref. [11]); (ii) frequency dependence: if
predators always choose prey of the most common color,
then a mutation for a new prey color will increase survival
and will therefore spread until it becomes common enough
that predators prefer it (e.g. Ref. [12]); (iii) timing effects: if
a population is growing, mutations for earlier reproduction
are successful because the earlier offspring are produced,
the more descendants they leave (discussed in Ref. [13]). In
all three cases, the (st)age trajectories of the dominant
strategy of a population determine population dynamics
and, thus, whether a new mutation can spread.
Although theoretical tools are available to handle
dependence of fitness on the population context (invasi-
bility analysis, Table 1), we are only beginning to under-
stand how such dependencies operate in natural systems.
For example, methods for estimating density dependence
in stage-structured populations have only just become
available [14] and more empirical work is a crucial next
step. Furthermore, any interdependence between popula-
tion- and individual-level outcomes leads to an inherently
dynamic situation (e.g. Ref. [15], reviewed in Ref. [10]).
Although adaptive dynamics approaches [16] have pro-
vided us with a novel and intuitive solution to analyzing
this problem, the invasion criteria most commonly used is
invasion when rare. The fate of an invader in a population
once it has successfully passed the status of rare is rarely
discussed, but might have complex implications for popu-
lation dynamics and evolutionary outcomes (e.g. Ref. [17]),
and be important for progress in the field.
Drift
Drift refers to the fact that, purely by chance, some
mutations will become ubiquitous in a population even if
they increase neither fertility nor survival, and even
slightly decrease them. The population context matters
for drift because the key variable driving this stochastic
process is the effective population size, Ne, which is influ-
enced by fluctuations in total population size and (st)age-
structure, as well as by variance in reproductive success.
These are accurately captured only by (st)age-trajectories
of survival and fertility. Theoretical work often makes the
assumption that generations do not overlap and many
www.sciencedirect.com
results are robust to this assumption [18]. However,
techniques for calculating Ne for overlapping generations
are available (reviewed in Ref. [19]) and resulting esti-
mates of the magnitude of genetic drift can double [20].
Sophisticated approaches have recently been developed to
incorporate the effect of individual differences in survival
and fertility on Ne via resulting fluctuations in age-struc-
ture and population size (e.g. Ref. [21]). A key question for
future research, generally neglected by evolutionary ecol-
ogy, is to what degree the evolution of age trajectories is
determined by genetic drift.
Dispersal
The spatial structure of populations might be as important
as (st)age structure, because selection pressures can
change across space. Combined with dispersal rates, this
might significantly alter evolutionary outcomes [22], for
example, enhancing [23], or diminishing [24] the coexis-
tence of different traits. Spatial connectivity is often poorly
understood, but increasingly sophisticated capturerecap-
ture techniques can provide sufficient information even for
linking alleles to demographic patterns [25]. (St)age spe-
cific dispersal is also likely to be the norm for most species,
particularly where juveniles establish new territories on
attaining maturity. For example, different passerine spe-
cies [26] and humans [27] migrate at different ages; (st)age
specific dispersal might modify population dynamics by
altering (st)age structure and affecting population growth,
with repercussions for selection and drift. Although optim-
ality models have been applied to the evolution of dispersal
(e.g. Ref. [28]), the (st)age specific dimension is rarely
considered and is ripe for development.
Long-term evolutionary outcomes: the diversity of
(st)age trajectories across species
Species display a variety of age trajectories. Life span
varies from a few hours (mayflies) to hundreds of years
(trees). Fertility also varies considerably (Table 2). Certain
broad-scale patterns, such as the slowfast continuum of
mammals (after removing the effect of size, mammals can
be categorized along a continuum from species with late
maturity, few offspring per reproductive event and a long
generation time, to species that reproduce early, have large
litters and a short generation time), are relatively well
understood [29]. However, many species deviate from these
general trends: for example, the small precocial mammal
Cavia magna produces few juveniles with high survival
when they are expected to produce large litters of
altricial juveniles [30].
Although the foundations of a cross-species
understanding of demography have been laid [31], several
major aspects remain unclear. For example, mortality in
many species increases with age (senescence), although it
can also remain constant (e.g. hydra [32]) or decrease (e.g. a
monocarpic plant [1]). Recent models have begun to
address such patterns. For example, Baudisch [33] showed
that the force of selection does not necessarily decrease
with age if fertility and survival increase sufficiently with
age. This might explain why some species do not senesce.
More generally, advances in mechanistically based theo-
retical demography (e.g. Ref. [34]) provide a framework for
 Review TRENDS in Ecology and Evolution Vol.22 No.4 209

Table 2. The diversity of age trajectories

Organism Average population Comment Ongoing research areas Refs
trajectory of survival
and fertility a
Humans Mortality is high for infants and is then low until maturity, What causes senescence [52,53]
corresponding to the initial drop in survival. In late (, mutation accumulation
adulthood, mortality increases gradually (senescence) or antagonistic pleiotropy
and survival continues to falls away. Maximum recorded or optimization of resource
age is 122. Fertility peaks at 30 and then falls to zero allocation)?
at 50 (the age of menopause) Why do humans have the
highest fertility among
primates?
Why do menopause and
post-reproductive life exist?
Hydra (lab Over a period of four years, mortality was low and How can a species escape [32]
population) showed no sign of increase with age; consequently, senescence even though
survival diminished gradually. Fertility (both asexual and (presumably) mortality in
sexual reproduction) remained approximately constant natural populations is high?

Species with Juvenile mortality is high; mortality then slowly How can mortality continue [34] b
indeterminate growth decreases in association with increasing size, so that to decrease (negative
mortality can continue to fall even after the start of senescence)?
reproduction. Fertility is also associated with increasing In the absence of density
size. Theoretically, growth could continue and size could dependence, can growth
increase indefinitely, but practically, growth is limited by and therefore fertility continue
resource availability, usually driven by density to increase indefinitely?
dependence
Monocarpic plantsc Mortality decreases with increasing size. Once the age Why do many monocarpic [1]
at flowering is reached, individuals flower and die plants delay reproduction?
How can broad ranges of
different flowering sizes or
ages coexist?

a
Survival (red curve; i.e. the proportion of individuals surviving to each age) and fertility (blue curve; i.e. the mean number of offspring produced by surviving individuals at
each age) on an arbitrary scale.
b
See Ref. [34] for a model including such dynamics based around rockfish, a group of species whose longevity varies between 12 and at least 200 years.
c
Image reproduced with permission from Barry Rice.

compiling data across the range of species to clarify the
underlying processes driving age-trajectory evolution. An
interesting direction for future research is collection of
demographic data on more species across the tree of life.
This would provide a deeper understanding of the dynamic
interplay between age trajectories and selection, and
would clarify the extent to which age trajectories are
constrained by phylogeny [35] (as yet unclear).
Going beyond aggregate (st)age trajectories
The most widely used fitness measures are based on
population averages (e.g. the net reproductive rate, R0;
the finite rate of increase, l; and the stochastic growth
rate, ls, defined in Table 1). However, in most popula-
tions, even in a constant environment, individuals differ
from one another, leading to variance in the number of
offspring produced over the life course. This variance has
two components: (i) the difference between individuals
that survive to reproduce and those that do not (this
variance will exist even if every individual in the popu-
lation is strictly identical and has an identical risk of
dying, and can be estimated using mean age trajec-
tories); and (ii) individual differences in fertility and
survival (some individuals might be more fecund, or
more robust than others).
www.sciencedirect.com
The second type of variation can arise for several
reasons: (i) differences might be a direct outcome of envir-
onmental variation. Some seeds might land in patches that
are more resource-rich than are others, entraining differ-
ences in age trajectories of survival and fertility. Such
variation will not have a genetic component in that
offspring might not at all resemble their parents; (ii)
individuals might respond adaptively to environmental
variation [36], altering their (st)age trajectories to best
match environmental conditions. In Drosophila [37] and
other species, lifespan can respond plastically to caloric
restriction: if resources become scarce, individuals reduce
metabolism, allocating resources preferentially to survi-
val; (iii) variation might persist because different (st)age
trajectories have equivalent fitness, enabling the long-
term coexistence of different genotypes. For example, some
individuals might allocate more to survival at the expense
of competitive ability, whereas others might suffer higher
mortality but monopolize resources and reproduce more
frequently (e.g. Ref. [34]); (iv) variation might persist
because, over the timescales considered, new mutations
have not had enough time to go to fixation or be driven to
extinction.
As a result of all four aspects listed above, natural
populations will be aggregates of individuals with different
210 Review TRENDS in Ecology and Evolution Vol.22 No.4
Box 1. Human demography and evolutionary biology
Demography is central to evolution, and it is therefore unsurprising
that human demography contributes to the study of evolutionary
biology. A first major asset is data availability: more information is
available on humans than on any other species. Data sets often
contain thousands of records, repeated measures on individuals,
considerable environmental information and even complete genea-
logies. Such data can be used to test evolutionary hypotheses, such
as the role of parental care in the evolution of post-reproductive life,
that could not be tested on other species. Considerable genetic data
are also available. Coupled with knowledge of human demographic
history (bottle-necks, migrations, etc.) genetic data enable the
quantification of gene propagation (e.g. Ref. [62]). Conversely,
genetic polymorphism and evolutionary models inform our under-
standing of human demographic and social history (e.g. Refs.
[63,64]).
Theoretical advances in classical demography in fields such as
period and cohort effects, unstable populations, migration, hidden
heterogeneity and so on, will also contribute as greatly to
evolutionary biology in the future as they have in the past. For
example, the demographic theory of unstable population dynamics
[65] might clarify evolution in non-stationary populations, an area
where ecological models are recognized as being inadequate [66].
Contributions might also be methodological, informing estimates of
age trajectories in unstable populations via information from two
census points [67] and so on.
If demography serves evolutionary biology, evolutionary biology
also contributes to demography. Failure to forecast population
patterns resulting from increasing longevity [68] led demographers
to turn to biology. Evolutionary biology provides specific predic-
tions on aging [33] and also contributes to our understanding of
diseases whose rampant effects are linked to rapid evolution, such
as influenza (e.g. Ref. [69]). Demographers and biologists are also
increasingly working together (e.g. Ref. [70]): a synthesis of
techniques and discoveries made independently in both fields will
successfully inform both in the future.
characteristics of survival and fertility. Why is this varia-
tion challenging? First, fitness outcomes might change
substantially. For example, Rees et al. [1] showed that
individual variation in growth changes the predicted opti-
mal flowering size in monocarpic plants (i.e. the fitness
estimate corresponding to the mean growth curve does not
accurately portray fitness for individuals within the popu-
lation). However, this challenge can be met if individual
variation in growth is specifically estimated. A more com-
plex example learnt from human demography (Box 1) is
how aggregate measures of survival can obscure indivi-
dual-level patterns. Average population survival curves
flatten off at advanced ages. However, this flattening might
not accurately reflect the survival trajectory of an indivi-
dual, but might be observed only because frail individuals
die early, so that only individuals with overall lower mor-
tality are present at late ages [38,39]. Understanding the
evolution of aging requires information about the relative
fitness of these two types of individual. However the frailty
attribute constitutes a hidden heterogeneity among indi-
viduals, which can never be directly measured, but is
important because survival is a dynamic process.
More generally, individual age trajectories of survival
and fertility can (co)vary in complex ways as a result of
direct connections between fitness components (reproduc-
tion can decrease survival) or indirect connections (growth
can affect both survival and fertility; e.g. Ref. [1]). Because
of the dynamic aspect of survival outlined earlier, a
www.sciencedirect.com
demographic perspective is essential to understand the
implications of this (co)variation for evolution. Recent
methodological advances, such as Hierarchical Bayes
approaches, can be used to partition and structure models
of variation in aggregates in a demographic framework
(e.g. Refs [40,41]). Advances in pedigree analysis also
enable heritable variation in (st)age trajectories to be
distinguished from environmental noise [42], but sufficient
data are only just becoming available. Allying such esti-
mation of individual demographic (co)variation with mod-
els for predictions of evolutionary outcomes is an area in
considerable development (e.g. integral projection models
[43]; or indirect approaches [44]), with many unanswered
questions, even at the most basic level. For example, it is
still not clear, even for humans, what fraction of variation
in age trajectories of survival is heritable, what fraction is
environmentally imposed and what fraction is due to
plasticity.
Environmental variation through time might not only
generate variation in the age trajectories of individuals,
but also impose selection pressures on this variation. For
example, in some cases, the more survival and fertility
vary through time, the more ls is decreased [45] (this
breaks down if environments are autocorrelated [46]).
Vital rates with the biggest effects on ls (as identified by
perturbation analysis) should therefore be buffered
against variation, as a result of selection, so that we might
expect fewer differences between individuals at any given
time. This is the case in some natural populations (e.g.
Refs. [2,47]). Another complication connected to heteroge-
neity and currently opening up a new field of investigation,
is that environmental conditions can lead negative
covariation between selection pressures and the genetic
variance that they will act on, so that little evolutionary
change is possible [48].
Having outlined challenges arising both from estimation
of individual (st)age trajectories and modeling heteroge-
neous populations, we now turn to cases where even accu-
rate estimates of individual (st)age trajectories might be
insufficient, another current key challenge.
Going beyond trajectories of survival and fertility
Within the framework of stable population theory, age
trajectories of survival and fertility enable calculation of
the (st)age-specific contribution of one (st)age class to the
total number of descendants. This provides us with the
tools to dissect evolutionary processes at the individual,
population and phylogenetic levels. However, these tools
are not sufficient if age trajectories of survival and fertility
are linked across generations. For example, offspring sur-
vival often depends on the age of the parents at childbirth.
This occurs in humans owing to an increase in germline
mutation rates with age in males [49]; or to increased risk
of a child with Downs Syndrome with age for females (of up
to 35% for ages 45) [50]; and has also been recorded for
Drosophila [51]. Generally, the result is a decrease in
offspring quality with age of parents.
The issue of linkage across generations also arises
where parental care is required for offspring to survive
and reproduce successfully. Adults can increase fitness by
investing time and energy in existing offspring instead of
 Review TRENDS in Ecology and Evolution
current reproduction. They thereby contribute to fitness
without reproducing, even during post-reproductive peri-
ods [52,53]. Although the age trajectory of this investment
towards offspring, and its relationship with offspring fit-
ness, are difficult to measure empirically and difficult to
model theoretically, the way that it impacts population
dynamics and evolution is the subject of increasing
research interest, and recent theoretical and empirical
advances [15,54,55].
To conclude, simple measures and models of fertility
are often not adequate to measure gene transmission:
distinctions must be made between offspring of parents
of different ages. Calculating the (st)age specific contribu-
tion to fitness of survivors at one (st)age class is therefore
complex and both perturbation analysis and optimization
are extremely challenging. This challenge is only starting
to be met.
Conclusions
Evolutionary outcomes result from an interplay between
ultimate and proximate determinants of variation in the
trajectories of survival and fertility across (st)ages, within
and between species, and owing to phylogenetic history.
Complexities, covariation and feedbacks abound at each
level, and must be addressed, as they can have broad
implications. A demographic framework is the only way
to do this. Demographic research has recently made giant
strides towards developing statistical models of heteroge-
neity across individuals. Much exciting work is resulting
from the integration of this progress into evolutionary
analysis. Every month, new tools appear, such as methods
enabling empirical quantification of density dependence
from time series of (st)age structured populations [14], and
new properties are clarified, such as those of stochastic
environment elasticities [9]. However, new questions
appear with equal regularity, such as the demographic
implications of parental care (e.g. Ref. [53]), or how adap-
tation at the individual level can interact with population
dynamics (e.g. Refs. [17,25]). Evolutionary demography is
in a period of intense activity and the current productive
focus on (st)age trajectories might improve the under-
standing for both evolutionary biologists and human demo-
graphers of key questions, such as the evolution of
senescence.
Acknowledgements
We thank the evolutionary demography group at Rostock, T. Coulson, H.
Caswell, D. Koons, A. Baudisch, and E. Cam for discussion and comments
on earlier drafts. We also thank S. Munch, B. Rice and the demography
group at Rostock for providing images.
References
1 Rees, M. et al. (2000) The effects of heterogeneity on dispersal and
colonization in plants. In The Ecological Consequences of
Environmental Heterogeneity (Hutchings, M. et al., eds), pp. 237
265, Blackwell Science
2 Morris, W.F. and Doak, D.F. (2004) Buffering life histories against
environmental stochasticity: accounting for spurious correlation
between the variabilities of vital rates and their contributions to
fitness. Am. Nat. 163, 579590
3 Coulson, T. et al. (2003) Estimating selection on neonatal traits in red
deer using elasicity path analysis. Evolution 57, 28792892
www.sciencedirect.com
Vol.22 No.4 211
4 Coulson, T. et al. (2006) Estimating individual contributions to
population growth: evolutionary fitness in ecological time. Proc. R.
Soc. B 273, 547555
5 Kingsolver, H.E. et al. (2001) The strength of phenotypic selection in
natural populations. Am. Nat. 157, 245261
6 Grafen, A. (1998) On the uses of lifetime reproductive success. In
Reproductive Success (Clutton-Brock, T., ed.), pp. 454471, University
of Chicago Press
7 Caswell, H. Sensitivity analysis for transient population dynamics.
Ecol. Lett. (in press)
8 Koons, D.N. et al. (2005) Transient population dynamics: relations
to life history and initial population state. Ecol. Model. 185, 283297
9 Haridas, C.V. and Tuljapurkar, S. (2006) Elasticities in variable
environments: properties and their implications. Am. Nat. 166,
481495
10 Benton, T.G. et al. (2006) Complex population dynamics and complex
causation: devils, details and demography. Proc. R. Soc. B 273, 1173
1181
11 Huber, N.C. et al. (2004) Frequency and microenvironmental pattern of
selection on plastic shade-avoidance traits in natural populations of
Impatiens capensis. Am. Nat. 163, 548563
12 Olendorf, R. et al. (2006) Frequency dependent survival in natural
guppy populations. Nature 441, 633636
13 Brommer, J.E. et al. (2002) Reproductive timing and individual fitness.
Ecol. Lett. 5, 802810
14 Lande, R. et al. (2006) Estimating density dependence from time series
of population age structure. Am. Nat. 168, 7687
15 Beckerman, A. et al. (2002) Population dynamic consequences of
delayed life-history effects. Trends Ecol. Evol. 17, 263269
16 Geritz, S.A.H. et al. (1997) Dynamics of adaptation and evolutionary
branching. Phys. Rev. Lett. 78, 20242027
17 Greenman, J. et al. (2006) The evolution of population dynamics in age-
structured species. Am. Nat. 166, 6878
18 Charlesworth, B. (1994) Evolution in Age-Structured Populations (2nd
edn), Cambridge University Press
19 Caballero, A. (1994) Developments in the prediction of effective
population size. Heredity 73, 657679
20 Rogers, A. and Prugel-Bennett, A. (2000) Evolving populations with
overlapping generations. Theor. Pop. Biol. 57, 121129
21 Engen, S. et al. (2005) Effective size of a fluctuating age-structured
population. Genetics 170, 941954
22 Ferrierre, R. and Le Gaillard, J-F. (2001) Invasion fitness and adaptive
dynamics in spatial population models. In Dispersal (Clobert, J. et al.,
eds), pp. 5779, Oxford University Press
23 Mathias, A. and Kisdi, E. (2002) Adaptive diversification of
germination strategies. Proc. R. Soc. B 269, 151155
24 Stratton, D.A. and Bennington, C.C. (1998) Fine-grained spatial and
temporal variation in selection does not maintain genetic variation in
Erigeron annuus. Evolution 52, 678691
25 Hanski, I. and Saccheri, I. (2006) Molecular level variation
affects population growth in a butterfly metapopulation. PLoS Biol.
4, e129
26 Woodrey, M.S. and Chandler, C.R. (1997) Age-related timing of
migration: geographic and interspecific patterns. Wilson Bull. 109,
5267
27 Rogers, A. et al. (2002) Capturing the age and spatial structures of
migration. Environ. Plan. A 34, 341359
28 Poethke, H.J. et al. (2003) Local extinction and the evolution of
dispersal rates: causes and correlations. Am. Nat. 161, 631640
29 Promislow, D.E.L. and Harvey, P. (1990) Living fast and dying young: a
comparative analysis of life-history variation among mammals. J. Zool.
220, 417437
30 Kraus, C. et al. (2005) Living slow and dying young? Life-history
strategy and age-specific survival rates in a precocial small
mammal. J. Anim. Ecol. 74, 171180
31 Charnov, E.L. (1993) Life History Invariants, Oxford University Press
32 Martinez, D.E. (1998) Mortality patterns suggest a lack of senescence
in Hydra. Exp. Gerontol. 33, 217225
33 Baudisch, A. (2005) Hamiltons indicators of the force of selection. Proc.
Natl. Acad. Sci. U. S. A. 102, 82638268
34 Bonsall, M.B. and Mangel, M. (2003) Life history trade-offs and
ecological dynamics in the evolution of longevity. Proc. R. Soc. Lond.
B. Biol. Sci. 271, 11431150
212 Review TRENDS in Ecology and Evolution Vol.22 No.4
35 Saether, B-E. and Bakke, O. (2000) Avian life history variation and
contribution of demographic traits to the population growth rate.
Ecology 81, 642653
36 Meyers, L.A. and Bull, J.J. (2002) Fighting change with change: adaptive
variation in an uncertain world. Trends Ecol. Evol. 17, 551557
37 Mair, W. et al. (2003) Demography of dietary restriction and death in
Drosophila. Science 301, 17311733
38 Vaupel, J.W. et al. (1979) The impact of heterogeneity in individual
frailty on the dynamics of mortality. Demography 16, 439454
39 Zens, M.S. and Peart, D.R. (2003) Dealing with death data: individual
hazards, mortality and bias. Trends Ecol. Evol. 18, 366373
40 Cam, E. et al. (2002) Individual covariation in life history traits: seeing
the trees despite the forest. Am. Nat. 159, 96105
41 Clark, J. (2003) Uncertainty in population growth rates calculated
from demography: the hierarchical approach. Ecology 84, 13491360
42 Charmantier, A. et al. (2005) Age-dependent genetic variance in a life-
history trait in the mute swan. Proc. R. Soc. B 273, 225232
43 Ellner, S.P. and Rees, M. (2005) Integral projection models for species
with complex demography. Am. Nat. 167, 410428
44 Weladji, R.B. et al. (2006) Good reindeer mothers live longer and
become better in raising offspring. Proc. R. Soc. B 273, 12391244
45 Tuljapurkar, S. (1990) Population Dynamics in Variable Environments,
Springer
46 Tuljapurkar, S. and Haridas, C.V. (2006) Temporal autocorrelation
and stochastic population growth. Ecol. Lett. 9, 327337
47 Gaillard, J.M. and Yoccoz, N.G. (2003) Temporal variation in survival
of mammals: a case of environmental canalization? Ecology 84, 3294
3306
48 Wilson, A.J. et al. (2006) Environmental coupling of selection and
heritability limits evolution. PLoS Biol. 4, e216
49 Crow, J.F. (2000) The origins, patterns and implications of human
spontaneous mutation. Nat. Rev. Genet. 1, 4047
50 Morris, J.K. et al. (2005) Risk of a Down syndrome live birth in women
45 years of age and older. Prenatal Diag. 25, 275278
51 Priest, N.K. et al. (2002) The role of parental age effects on the evolution
of aging. Evolution 56, 927935
52 Peccei, J. (2001) Menopause: adaptation or epiphenomenon? Evol.
Anthropol. 10, 4357
53 Pavard, S. et al. The effect of maternal care on child survival:
a demographic, genetic and evolutionary perspective. Evolution
(in press)
Forthcoming Conferences
Are you organizing a conference, workshop or meeting that would be of interest to TREE readers? If so, please e-mail
the details to us at TREE@elsevier.com and we will feature it in our Forthcoming Conference filler.
79 May 2007
The Mathematics of Evolution: a Fields Institute
Workshop on Adaptive Dynamics in Theory and
Practice, University of Ottawa, Canada
http://www.science.uottawa.ca/~fluts037/FIELDS/
fieldsworkshop.html
1012 May 2007
Biological Sciences for the 21st Century: Meeting the
Challenges of Sustainable Development in an Era of
Global Change, Washington, DC, USA
http://www7.nationalacademies.org/IUBS
1319 May 2007
VIIIth Neotropical Ornithological Congress, Estado
Monagas, Venezuela
(http://www.esa.org/federation/meetings/
FirstCallContributions_VIIINOC.pdf)
2227 May 2007
EcoSummit 2007: Ecological Complexity and
Sustainability: Challenges and Opportunities for 21st-
Centurys Ecology, Beijing, PR China
http://www.ecosummit2007.elsevier.com
www.sciencedirect.com
54 Kaplan, H.S. and Robson, A.J. (2002) The emergence of humans: the
coevolution of intelligence and longevity with intergenerational
transfers. Proc. Natl. Acad. Sci. U. S. A. 99, 1022110226
55 Chu, C.Y. and Lee, R.D. (2006) The co-evolution of intergenerational
transfers and longevity: an optimal life history approach. Theor. Pop.
Biol. 69, 193201
56 Caswell, H. (2001) Matrix Population Models: Construction, Analysis
and Interpretation (2nd edn), Sinauer Associates
57 McGraw, J.B. and Caswell, H. (1996) Estimation of individual fitness
from life-history data. Am. Nat. 147, 4764
58 Link, W.A. et al. (2006) Model-based estimation of individual fitness. J.
Appl. Stat. 29, 207224
59 Grant, A. and Benton, T.G. (2000) Elasticity analysis for density-
dependent populations in stochastic environments. Ecology 81, 680693
60 Mylius, S.D. and Diekmann, O. (1995) On evolutionarily stable life-
histories, optimization and the need to be specific about density-
dependence. Oikos 74, 218224
61 Brommer, J.E. et al. (2004) Single generation estimates of individual
fitness as proxies for long-term genetic contribution. Am. Nat. 163,
505517
62 Austerlitz, F. and Heyer, E. (1998) Social transmission of reproductive
behavior increases frequency of inherited disorders in a young-
expanding population. Proc. Natl. Acad. Sci. U. S. A. 95, 1514015144
63 Chaix, R. et al. (2004) The genetic or mythical ancestry of descent
groups: lessons from the Y-chromosome. Am. J. Hum. Genet. 75, 1113
1116
64 Bertranpetit, J. (2000) Genome, diversity, and origins: the Y
chromosome as a storyteller. Proc. Natl. Acad. Sci. U. S. A. 97,
69276929
65 Keyfitz, N. (1968) Introduction to the Mathematics of Populations,
Addison-Wesley
66 Lande, R. et al. (2003) Stochastic Population Dynamics in Ecology and
Conservation, Oxford University Press
67 Preston, S.H. and Coale, A.J. (1982) Age structure, growth, attrition,
and accession: a new synthesis. Pop. Index 48, 217259
68 Oeppen, J. and Vaupel, J.W. (2002) Broken limits to life expectancy.
Science 9296, 10291031
69 Earn, D.J.D. et al. (2002) Ecology and the evolution of flu. Trends Ecol.
Evol. 17, 334340
70 Keyfitz, N. and Caswell, H. (2005) Applied Mathematical Demography
(3rd edn), Springer
58 June 2007
ESEE 2007. Integrating Natural and Social Sciences
for Sustainability, Leipzig, Germany
http://www.esee2007.ufz.de
1215 June 2007
5th Meeting of the Australasian Evolution Society,
The University of New South Wales, Sydney, Australia
http://www.evolutionau.org
1621 June 2007
Evolution Annual Meeting, held jointly by the Society
for the Study of Evolution, the American Society of
Naturalists and the Society of Systematic Biologists,
Christchurch, New Zealand
http://www.evolutionsociety.org/meetings.htm
15 July 2007
Society for Conservation Biology Annual Meeting, Port
Elizabeth, South Africa
http://www.nmmu.ac.za/scb/
1618 July 2007
ENTO 07: RES Annual National Meeting and RES
Symposium on Aquatic Insects, Edinburgh, UK
http://www.royensoc.co.uk/