Charles Darwin
1809-1882
1
The basic framework of natural selection as constructed by C.D. (published in 1859,
nearly 20 years after his trip around the world):
1. IF entities reproduce
2. AND there is heredity (offspring are more like parents than ‘average’ properties of
the population,
3. AND there is variation among individuals in terms of heritable traits
4. AND populations tend to grow so long as conditions permit (Malthus and
exponential growth)
5. AND resources are finite (OR the world somehow constrains numbers)
6. THEN there will be competition (OR some form of limitation on a growing
population)
7. AND it is likely that individuals possessing one form of a variable phenotypic
‘trait’ will be BETTER AT CAPTURING resources or dealing with constraints
8.
SO THAT they’ll be more likely to reproduce than individuals possessing other
forms of the trait
9.
AND that more ‘adaptive’ trait will increase in frequency until it replaces others,
changing the genetic make-up of the population at issue
1
Charles Darwin
1809-1882
2
SO: Natural selection is simply changes in frequency (proportional abundance) of
heritable phenotypic traits (or of homologous alleles for a gene) due to the effects of
those traits or alleles on the individual organism’s reproductive success compared to
other individuals in a population.
FIVE IMPORTANT THINGS TO REMEMBER:
- It’s driven by differences among individuals and operates on individuals (not
properties of species or populations)
- But what EVOLVES/changes is the genetic make-up of POPULATIONS.
- It works ONLY on variation that is available within the population; it can’t make
new variations and the ‘best’ possibilities may not occur
- It works ONLY through differences in contribution of relatives to future generations;
other things (like survival) only matter if they influence this.
- It is driven only by CURRENT CONDITIONS; there is no means by which the
process can work towards FUTURE benefits
"Natural selection will operate wherever there is heredity, variation (and that will
almost inevitably follow given you have got heredity), competition (and that will
almost inevitably follow given you have got heredity and variation). So I suspect that
the conditions for natural selection to work are very minimal indeed: namely, the
existence of the phenomenon of heredity.“ -- Richard Dawkins
2
3
Darwin’s Finches: Distributions of frequencies of beak sizes represent one measure of
fundamental niche – the range of seed sizes and shapes finches CAN exploit. Note
differences in beak-size distributions for Geospiza fortis among populations on various
islands. Where it co-occurs with smaller-beaked G. fuliginosa, beak sizes are larger.
We know that beak-size is essentially 100% heritable (genetically determined), so this
is a genetic difference, presumably due to evolutionary divergence between
populations.
Likely scenario: on an island where the two species are ‘sympatric’ (co-occurring),
larger fortis individuals suffer less from competition with fuliginosa and so reproduce
more abundantly than their smaller-beaked con-specifics (REMEMBER that selection
is about differences between individuals within the same population). Over time, then
population becomes dominated by big-beaked birds (or, alternatively, gene-pool
becomes dominated by big-beak alleles). Because this process involves selection
favoring individuals at one extreme of a distribution, it tends to shift the whole
distribution of the property towards that extreme; this is called directional selection.
Note, also, that this scenario produces a change in the fundamental niche of the birds;
there are no longer G. fortis with really small beaks, so no members of this species are
likely to be able to handle really small seeds regardless of the presence of competitors
(before the effects of directional selection, small-beaked fortis may have been poor
competitors for small seeds so the realized niche might not have included much use of
small seeds IF competitors were present.)
3
4
Darwin’s Finches continued: NOW, think about:
a) The narrower distribution of fortis beak sizes in the top graph where a largerbeaked species, G. magnirostris, is present. This might result from competition
disfavoring genes for big beaks AND genes for small beaks. This would amount
to stabilizing selection, which tends to eliminate extremes of initial distribution,
reducing overall variation. (Disruptive selection is when the most frequent
phenotype is less favored by selection than rare phenotypes; in this case selection
would tend to maintain or increase genetic variation within a population; think
about when this might happen).
b) FINALLY, think about a historical selective scenario that might explain difference
between top two lines.
4
5
A parallel scenario for two species of land snails living on islands. Upper left graph
shows shell-lengths (mean and range) for allopatric populations (i.e., population on an
island without the other species). Open symbols are one species, closed symbols the
other. Lower left graph shows a series of island where both species occur in
sympatric. In all cases, H.ulvae now has significantly larger shells and H. ventrosa
smaller shells than on islands where either occurs alone, presumably due to directional
selection due to competition between them. This phenomenon of competition
selectively driving directional shifts in distribution of a character (as with the finch’s
beaks) is called character displacement. Again, it is a genetic/heritable property that’s
changing and the consequence is a change in the fundamental niche of the species.
5
6
Ants in the Sonoran Desert of Arizona and California. Top left: Each dot is a
population of ants. ‘CV’ or coefficient of variation is an index of variability; in this
case larger CV means mandible length varies more within that population. This graph
shows that populations in more diverse communities of seed-eating ants have less
variable mandible lengths – they are more uniform in this trait. This is consistent with
competition tending, over evolutionary time (generations) acting as an agent of
‘stabilizing selection’. Lower right graph shows mandible size distributions for a
single species in a number of populations. In each case, arrows show average
mandible size for competing species at the same site.
6
7
FITNESS is a central concept in models of evolution by natural selection. It is strictly
defined as the relative reproductive success of different individuals or genotypes
within a population. It is NOT the same thing as ‘adaptation’ which is a measure of
how well-engineered a phenotype is to environmental circumstances. Differences in
adaptation may certainly influence fitness, but fitness must ALWAYS and ONLY be
defined only in terms of differences in reproductive success. A genotype or allele that,
on average, leaves more descendants than other genotypes/alleles, has highest fitness
of available genotypes/alleles. An individual that has more offspring than another
generally has higher individual fitness (there are fine points here that we won’t get into
right now). And so on. Here, clutch size, the number of eggs per brood/nest/female,
for a European song-bird is shown to affect fitness (we know clutch size to be strongly
genetically determined – i.e., heritable). Birds with clutches of 9 or 10 have higher
fitness (here measured as number of known survivors from the clutch at 3 months of
age) than birds with either larger or smaller clutches. (NOTE that this assumes a
number of things – e.g., that mature females with different clutch-size-genes have
similar survival over years, that survivorship at 3 months is indicative of long-term
survival, etc.). This is a case of stabilizing selection. BUT CONSIDER A PROBLEM
HERE: If this selective scenario continued to play out, after a very few generations, the
genes for smaller and larger clutches would be ‘selected out’ of the population! Why
are there still such variants in the population? Hint: Think about WHY larger and
smaller clutches would confer reduced fitness, and then think about how selective
factors might change over time…
7
8
Evolution by natural selection in an ‘artificial’ digital life-form might be used as a
model to better understand evolution and adaptation in real life forms Karl Niklas – a
paleobotanist, used the shapes of very early land plants (see next page) to come up
with simple, branching ‘plant-like’ shapes defined by four shape parameters. These
are 1) frequency of branching (or distance between branches), 2) angle of bifurcation
at each branching event, 3) rotation of branching plane from one branching event to
the next, and 4) total number (or length) of branches before growth stops. There were
some more complexities in the end, but this is the basic plan. At the tip of each ‘final’
branch, the ‘plant’ produces a reproductive organ (spore-case in his model).
Now, think of each of these growth/shape parameters as a ‘gene’. For any individual
‘plant’ it has some numeric value. Let your plant reproduce – but let the values for
these genes mutate a little bit with some low probability (most offspring don’t mutate,
and mutations are always small). Now ‘create’ your plants in the computer as a few
lines of computer code written to generate shapes determined by the values of these
parameters.
8
9
Right panel is an actual fossil of one of the earliest land plants; left panel is a
reconstruction of what it might have looked like. They were only a few cm high.
These shapes could easily be generated by the computer simulations Niklas wrote.
Now consider adaptive properties of these plants/shapes. Niklas considered four ways
in which he could assess the ‘quality’ or ‘adaptedness’ of a shape:
1) light-capture effectiveness: simply the total surface area facing upward (‘projected’
area);
2) Water-use efficiency: since plants lose water through surfaces, Niklas simply
calculated total surface area (all surfaces). Lower surface area equals higher wateruse efficiency
3) Structural strength: Assuming constant properties of the materials used to build the
shapes, strength (resistance to being blown over) is a simple engineering
calculation based on mass, angles of branches, height, etc.)
4) Reproductive success: Niklas assumed wind-blown spores and based reproductive
success on the number of spore-cases (stem-tips) and how high they were (for
good dispersal).
All obviously simplistic, but remember that ‘modeling’ and theory approaches try to
start with simple models and see how realistically they behave, making them more
complicated as needed to make behavior sufficiently ‘real’.
9
10
Imagine all the shapes that could be produced by varying just two of the ‘genes’; these
can be arranged on a two-dimensional space/plane as at top; a few of the shapes
corresponding to values of ‘genes’ are shown as example.
Using the ‘adaptedness’ measurements (previous slide), each shape (combination of
two gene values) can be assigned a synthetic adaptation score; treat this, for now, as
equivalent to fitness.
Now, start with simplest shape (lower left). Allow it to reproduce and mutate, but
‘mutations’ can only change either gene by one step, so ‘offspring’ shapes are never
more than one square from the parent.
Each generation, choose the ‘offspring’ with highest fitness to produce the next
generation. Over time, the descendant population of shapes will tend to move, one step
at a time to, combinations of gene-values with higher fitness. Niklas calls this a ‘walk
through fitness space’. (Here, a walk over a fitness ‘surface’; think of the plane as a
topographic map with values representing elevation; selection will tend to move the
‘population’ up-slope.)
Note important correspondences to selective processes: a) Selection can only act on
variations available – there may be ‘better’ combinations out there, but we’re only
allowed to move one step at a time. B) there’s a chance element that determines which
‘branch’ history might take; it depends on which mutation happens, c) Once
population finds a ‘peak’ – a square surrounded by lower values – it is ‘fixed’ by
stabilizing selection. SO different outcomes happen in different runs, and there may be
several ‘peaks’ to which population CAN evolve (and not all of them are equally
‘good’!)
10
11
Same sort of concept in three dimensions – using three of the ‘genes’ for shape. Just
generalize to 4-gene space – or as many dimensions as there are genes for any type of
organism.
11
12
So, Niklas simulates walks through ‘morphospace’ – the universe of possible shapes –
with ‘fitness’ for each shape calculated on the basis of some combinations of the
adaptive aspects discussed earlier.
12
13
If fitness is based on just ONE of those adaptive tasks, these are the ‘peaks’ achieved
by selection acting in morphospace – there may be better designs, but they can’t be
reached. In three selective scenarios there are multiple ‘local adaptive peaks’ in
morphospace. Selection for spore dispersal alone, however, always leads to the same
shape.
13
14
These are the shapes produced by adaptive ‘walks’ in fitness landscapes defined by
combinations of two of the adaptive needs. Note that combining scores for mechanical
stability and spore dispersal again always leads to convergence on one shape – the
same shape as spore-dispersal selection alone. All other combinations of ‘tasks’
produce multiple ‘solutions’ that are better than any small variations on them (i.e.,
local optima in the ‘fitness space’). Combining light interception task with either
mechanical stability (E) or spore dispersal (F) tends to produce relatively realistic treeshapes…
14
15
Same thing combining THREE adaptive tasks in calculating fitness scores
15
16
And, finally, combining all four adaptive tasks. Now there are quite a few ‘local
peaks’ on the landscape – about 20 stable designs. Consider how this model
successfully reproduces ‘real-world’ patterns and where it doesn’t. What does all this
mean? Does it say anything about why there are diverse types of trees? Why there’s
no single ‘best-design’ tree that out-competes all others? An important concept to
consider is notion of TRADE-OFFS. Might doing one thing make you inherently less
effective at doing another?
16
17
Real trees
17
18
19
20
All of the ‘tree-things’ in previous example evolve by incremental steps – the radically
different end points are only achieved by accumulation of small changes over many
generations. Some critics have argued that complex structures can’t come about by
small changes because ‘intermediate’ forms would not work. But remember that the
‘intermediate’ forms only have to work better than anything else that’s around at the
time; they aren’t required to work as well as the current ‘end-product’. Eyes are often
held up as such a ‘challenge’ to evolutionary theory; ‘half an eye’ would not work,
claim the critics. But, in fact, a poorly functioning eye (half an eye) is a lot better than
none. These are all forms of light-sensing organs that occur; they form an incremental
series from a patch of pigmented cells to a fully functional vertebrate eye. (Note that a
simple light-sensitive ‘pit’ as in (b) will allow sensing of the direction from which
light is coming).
20
21
A classic real-world analysis of fitness: Moths roost on tree trunks during the day. In
Britain before industrialization, tree trunks were lichen-covered (top photo), and pale,
mottled moths were dominant, with only very rare ‘melanic’ (uniformly dark) mutants
(they were highly valued by collectors!). With industrialization, air pollution killed
lichens leaving tree-trunks dark gray-brown. Dark-colored moths rapidly increased to
dominance in the population. This is consistent with the notion that predation imposes
differential selection on these two phenotypes; the hypothesis would be that more
camouflaged individuals survive predation to reproduce with greater frequency.
Kettlewell tested part of this hypothesis by releasing marked moths of both types in
both polluted woodlands and unpolluted ones (where trees were still lichen-covered)
and then recapturing as many as possible; in both cases, recapture rates were higher for
the genotype predicted to be better able to avoid predation
21
22
Insect’s quickly evolve resistance to insecticides (toxins are powerful selective
factors!). These data are measured reproductive output of mosquito genotypes
resistant to two insecticides (fitness of the non-resistant, ‘wild’ genotype, is set at 1, so
all of these values are proportional to that). Resistant genotypes show much higher
fitness – reproductive rates often several times higher -- in environment with
insecticide (not surprising). BUT, in environments without the insecticide, their
reproductive output – or fitness – is actually lower than that of the wild type. This
suggests that insecticide resistance involves some form of trade-off. This is a general
pattern. Think about consequences for evolution of populations with and without
exposure to insecticide.
22
23
COEVOLUTION – when the ecological interactions between species affect the fitness
of each, so that each defines an important part of the selective environment of the other
– can lead to complicated and interesting dynamics. Indian-pipe is a flower plant that
lacks chlorophyll; it acquires sugars from a fungal associate rather than from its own
chlorophyll. There is no evident benefit to the fungus which is mycorrhizal – that is, it
lives in association with tree roots and draws its own sugars from green plants. So
Indian-pipe is parasitic on the fungus-greenplant mutualism. Consider why selection
would predictably lead to selective loss of chlorophyll in such a case. But shouldn’t
there be selection on the fungus to somehow ‘reject’ the Indian-pipe’s parasitism? Or
is its draining of resources too trivial in relation to the availability of sugars from the
tree roots to matter much? Or is Indian-pipe somehow able to evade any ‘rejection’ by
the fungus (maybe it looks too much like the roots of photosynthetic plants for the
fungus to differentiate? Action of selection in the interplay between parasites and
predators and their hosts or prey is an intensive area of study.
23
24
Plant-pollinator systems are generally thought of as mutualisms – both parties
experience positive, population-level effects. However, selection typically acts on
each party to minimize the costs to that party and maximize the benefits. Nectar is
expensive, so selection should favor plants that produce the minimum amount
necessary to ensure pollination. Selection should act on the ‘pollinator’ to maximize
foraging efficiency for nectar regardless of effectiveness with which it pollinates the
plant (which would have no effect on ‘pollinator’ fitness). Indeed, many types of
insects have evolved means of accessing nectar by means that don’t pollinate the plant
at all (you may be able to see the small holes chewed in the upper lobes of some of
these flowers; that’s where the nectar glands are located)
24
25
And some plants don’t reward the pollinator in any way at all; dark brown or redbrown flowers often emit the odor of decaying flesh and attract flies that normally lay
eggs on dead animals. There is no food for the fly (or it’s larvae); if the fly actually
lays eggs at the flower (they do in some cases), this is actually a fitness cost since
those offspring die. The plant is a parasite on the insect. In this case, it’s easy to see
the selective benefit to the plant (scents are cheaper than nectar; more energy left for
growth and seed production). But it seems that selection ought to favor insects who
can be discriminating and avoid this cost. Why do the flies still come to the flower?
25
26
This is an orchid that mimics a female bee, even emitting sex pheromones. Male bees
copulate with the flower and transfer pollen in the process. Again, the plant is
parasitic on the insect. Are these kinds of relationships likely to be stable – i.e., would
selection favor their maintenance? There are lots of similar stories.
26
27
And, finally, some illustrations of how the mechanism of selection leads to imperfect
and/or unexpected or odd results. The ‘panda’s thumb’ – actually an enlarged wrist
bone – is not a very ‘good’ thumb (no joints, not very strong or flexible), but pandas
with a poor excuse for a thumb can still strip bamboo faster than pandas without – and
there is no variation within the panda gene pool that allows selection to favor
‘opposability’ of one of the true fingers.
27
28
The sickle-cell trait – a simple recessive mutation of the hemoglobin gene in humans –
causes a tendency to anemia by causing red blood cells to be sensitive to stress (they
collapse into ‘sickle’ form and become nonfunctional). In homozygote condition (two
copies of the gene), carriers almost always die in childhood. In heterozygote form
(one copy of the mutation, one ‘normal’ copy of the gene), the tendency to anemia is
relatively mild, but the carrier is resistant to malaria, a disease that has killed vast
numbers of people before reproductive age.
28
malaria in southern Europe,
southwest Asia, and Africa
around 1920
sickle-cell allele within the same
area. The darker the blue,the
greater the percentage of people
carrying the allele.
29
Despite its great fitness cost to carriers – on average, a quarter of the children of two
‘carrier’ parents will die in childhood due to sickle-cell anemia – the sickle-cell gene is
prevalent throughout the region where malaria is endemic. Reduction of one large
fitness cost more than compensates for the other. A design trade-off that illustrates
that natural selection is not always an ‘efficient’ designer -- a good design would be
resistance without such a huge trade-off – but selection can only work with the
variations available from mutation.
29
30
The human appendix is a costly organ; it’s not functional in humans, but can cause
sickness and death if infected. The homologous organ functions as a digestive aid in
other mammals that eat low-quality foliage. Selection in humans, specialists in ‘highquality’ food (fruits and meat), has reduced the organ to a ‘vestigial’ state. Why isn’t
it altogether eliminated, reducing the selective costs of appendicitis? Perhaps it’s
because an even smaller appendix actually increases risk of appendix – so any small
change in appendix size leads to lower fitness (stabilizing selection), even though a
large change (to no appendix) would be great…
30
31
Even though elegant adaptations are abundant (woodpeckers – organic jackhammers),
they can’t be counted on. Where there are no ‘true’ woodpeckers, other birds exploit
some of the same resources, but much less efficiently. Selection can’t always be
counted on to make elegant adaptation. Again, the variations that provide the selective
path to ‘woodpecker-ness’ may not have been available in the right sequence.
31
32
Hummingbirds (new world only) are convergent with sunbirds (old world) – but there
are differences; sunbirds can’t hover.
32