Transcript Chapter 3 Methods

Research Methods in
Behavioural Ecology
Hypotheses, models,
predictions, & theories
• Principles are logical construction
– Hypotheses are too, but they’re more tentative
• Model: A formalized representation of a system
– Can be ‘hypothetical’
– Mathematical, computational, physical
• Prediction: A logical outcome of a model or
hypothesis that can be tested against data
• Studies: Attempts to disprove hypotheses
• Theory: A broadly general body of scientific
understanding
Correlation and Causation
• What is correlation? 3
• Height and traffic 2.8
tickets
2.6
• Phones and
2.4
longevity
2.2
• Confounding
2
variables
1.8
– Age, economic
development
– Drawing time
1.6
1.4
-4
-2
0
2
4
The logic of experiments
• Manipulation of
independent
variable
• Random
assignment
• Control groups
• Solves the
confound problem
– Drawing time
• Barn swallows
Natural Experiments
• “Nature” changes
independent
variable of interests
From Gil da Costa et al. 2002
–E.g., disasters,
human interference
• Harpy eagle calls
and howler
monkeys
http://www.dkimages.com/discover/previews/1453/11181132.JPG
The Phenotypic Approach
• The most common
approach
• “…we assume that
studies at the level of
the phenotype are
sufficient for identifying
the selective pressures
that exert themselves
on the organism…”
Phenotypic: Optimization I
• Optimality models
– Assumption: Behavior serves to
maximize some currency of fitness
• Energy intake, access to eggs, etc.
– Economic logic
• Behavioral choices incur fitness costs and
benefits; find maximum benefit – cost
• Allow us to predict what animals should do
if they are perfectly adapted to their
environments
Phenotypic: Optimization II
• But…
– Tradeoffs and constraints limit ‘perfectability’
– Our models should recognize that
optimization is maximization under constraints
– Proximately, animals behave to achieve
“objective function”
• Match of “cost function” to “objective function”
• What happens when the environment changes?
• An example of optimality thinking
Phenotypic:
Game Theory
• Employed when the cost function
is affected by the strategy played
by others in the population
– E..g, Which side to drive on, the sex
ratio game
• All players make adaptive choice
• Population level vs. individual fitness
– The evolutionarily stable
strategy (ESS) cannot be
invaded by one or a small
group of actors playing an
alternative strategy
http://www.abc.net.au/reslib/200707/r162700_598492.jpg
The Hawk-Dove Game
• Resource value = V
• Two strategies: Hawk & Dove
– Hawks attack everyone
• 50% chance of winning against hawk, 100% against dove
• Cost of losing a fight = L
– Doves display to doves, flee from hawks
• 50% chance of winning against dove, 0% against hawk, but
no cost of fighting
• Cost of display = D
Payoff to
Against Hawk
Against Dove
Hawk
(V-I)/2
V
Dove
0
V/2 - D
The Hawk-Dove Game
Payoff to
Hawk
Dove
Against Hawk
(V-I)/2
0
Against Dove
V
V/2 - D
• Suppose V = 50, L = 100, D = 10
Payoff to
Hawk
Dove
Against Hawk
-25
0
Against Dove
+50
+15
• With these parameters, it’s good to be the rare
phenotype
• A “Mixed ESS”
• This logic can explain the persistence of behavioral variability
• If V > L, hawk is usually a pure ESS
Proxies of Fitness
• The phenotypic approach assumes
equivalent demographic & genotypic fitness
• Proxies of demographic fitness
• Do these actually
relate to fitness?
– E.g, in the case
of chicks in the
nest, is there
assurance of
paternity?
Phenotypic engineering
• The logic of engineered
“mutation”
• Ablation experiments
– Muting Scott’s seaside sparrows
• Can show current stabilizing,
neutral, or directional selection
– How?
– What about selection in the past?
• Why might a trait be expressed
at a suboptimal level?
Limits to the Phenotypic
Approach
• Assumption of ‘state of adaptedness’
– What if environments are rapidly changing?
• Assumption of single locus, haploid
inheritance
– What about heterozygote advantage?
• Could coexistence of three “malaria” types be
explained from phenotypic observations?
• Assumption of unconstrained adaptation
– http://www.youtube.com/watch?v=enrLSfxTqZ
0&feature=related
The Genetic Approach
• Monogenic determinism of
behavior is rare
– for gene in fruit flies
• Polygenic determinism is
predominant
– Meshing gene / phenotype
relationships to defined traits
• Variable expression of
complex behavioral traits is
attributable to variation in
genes, the environment, and
epigenetic factors
Sokolowski 2001, Nat Rev.
Quantitative genetics of
behavior
• Comparing populations
– Common garden
experiments
• Controlling for simple
maternal effects
• Frightened fish
– Transplant experiments
• Looking for maternal
effects
– Crossing experiments
Brown et al. 2007 BES
Quantitative genetics of
behavior II
• Artificial selection
–Continuous traits
–Generally, diversifying selection
–What do these kinds of
experiments tell us?
• What can be learned from
cloning? Knockouts?
The Comparative
Approach
• Comparing traits among species
• Without phylogeny
– Relate behavior to ecological factors across
spp. or populations
• E.g., Eggshell removal
– Problems
• Causation?
• “Cherry picking” examples that support hypotheses
• Confounds, like size, phylogeny
Allometry
• Body size is an important confound in comparative
studies
• Scaling one body part against another is tricky
• Allometry is the study of the relationship between
body measurements
• log(Y)= b log (X) + log (a)
• Slope (b) > 1 means Y
increases faster than X
– “positive allometry”
• Comparing residuals is
informative
Controlling for phylogeny
• Phylogenetic inertia
• Homoplasy and
homology
• Determining
ancestral characters
–Maximum parsimony
• Problem of equal
parsimony
Method of Independent
Contrasts
• Looks for relationship between
two continuous variables while
controlling for phylogeny
–Examples
• Assumes random change,
independent changes in
different branches
Is evolution of x correlated with
evolution of y (and if so, how)?
Method of Maximum
Likelihood
• Discrete variables
– E.g., duetting and monogamy
• 1st model: State changes in two variables
are independent (a)
• 2nd model: State changes are
interdependent (b)
• Can find most likely
direction, order