Presentation on theme: "The Elephant in the Dark:"— Presentation transcript:
1The Elephant in the Dark: Resolving evolutionary patterns across timescales
2One by one, we go in the dark and come out …One by one, we go in the dark and come outSaying how we experience the animalOne of us happens to touch the trunk.“A water-pipe kind of creature.”Another, the ear. “A very strong, always movingBack and forth, fan-animal.”Another, the leg. “I find it still,Like a column on a temple.”Another touches the curved back.“ A leathery throne”Another, the cleverest, feels the tusk.“A rounded sword made of porcelain.”He’s proud of his description.Each of us touches one placeAnd understands the whole in that way.The palm and fingers feeling in the dark areHow the senses explore the reality of the elephantIf each of us held a candle there,And if we went in together,We could see it.-The Elephant in the DarkRumi (Translated by Coleman Barks)
3Another, touches extant populations,… “A very strong, always moving back and forth [process]”
4Another, the fossil record, … “I find it still, like a column on a temple [with occasional punctuations]”
5Another, the phylogeny… …Brownian Motion might be a good approximation.
6How can the evolutionary pattern be all these things? Contemporary field studiesRapid evolutionHigh genetic variationStrong selectionFossil recordStasis over millions of yearsGeographic variationExtinction and species turnoverComparative methodsVariance increases through timeHigh morphological diversityBrownian motion fits reasonably
7How can we see the pattern across scales of time? All studies of phenotypic evolution measure comparable quantitiesAllochronicSynchronicTime intervalPop Amean(z)Pop Bmean(z)Pop Amean(z)Pop Bmean(z)Time intervalPop XWe measure two quantitaties:(1) “time for evolution”(2) Δ mean
8Divergence isn’t dependent on time? Estes and Arnold 2007HistoricField StudiesFossil RecordDivergence isn’t dependent on time?
9Why is there no time-span effect? How can this be consistent with strong phylogenetic signal for body size traits?Thomas F HansenUniversity of Oslo
10Body size is known to have a strong phylogenetic signal Smith et al Phylogenetic heritability of body size between sister species of mammals(think of the plot like a sib-sib regression)
11Divergence isn’t dependent on time? Estes and Arnold 2007HistoricField StudiesFossil RecordDivergence isn’t dependent on time?
12Estes & Arnold (2007) Neutral drift in trait or adaptive optimum Stabilizing selectionDisplaced optimumMoving optimumWhite noise motion of the optimumPeak shift
13Estes and Arnold 2007Well, the biggest compilation of data to date says no. Estes and Arnold compiled Field, Historic and Fossil data and plotted time measured in generations against divergence. They find no support for Brownian motion, largely because of this surprising result: The amount of divergence expected over 10 generations is essentially the same as that expected over 10 million. That is, there is no relationship between the evolutionary time separating populations, and the expected divergence.
14So how are we going to visualize this So how are we going to visualize this. We can easily just plot “time for evolution” on the x-axis, and on the y-axis the metric we are going to measure the change in mean by is the mean of the log-scaled measurements, so that divergence is a measure of the proportional change in a linear trait measurement. What do we expect evolution to look like?Time (Raw timescale)
15Example: Brownian Motion process If evolution looks like brownian motion, then we expect lineages to look something like this, where if there is no time between populations, we expect no divergence, and the longer the period of “evolutionary time” between populations, the larger the divergence.Time (Raw timescale)
16Example: Brownian Motion process Being a nicely behaved statistical model, we can put confidence limits on that. But notice that we are not looking “across” time here, as microevolutionary divergence occupies a vanishingly small sliver of this graph. So we need to go to logged-scaled time.Time (Raw timescale)
17Example: Brownian Motion process And this is the shape of Brownian motion on the log scale. So is that what the elephant looks like?Time (Log timescale)
18Comparative dataHistoricField StudiesFossil Record
19Let’s combine with comparative data: Log-scaled linear body size traitse.g. ln(height) or ln( mass1/3)Taxa- Many, but mostly vertebratesField and historic data(e.g. Hendry et al. 2008)Fossil record(e.g. Gingerich 2001)Comparative data:Time-calibrated phylogenies & body size databasesMammals, Birds & SquamatesSo let’s try again. This time we’re only going to look at a single class of traits, linear body size measurements. We’re going to look at all taxa together, but we’re also going to subdivide these out to look at individual groups. As in Estes and Arnold, we combine field and historic data, along with data from the fossil record. However, we expand this dataset considerably by including data from comparative methods. That is by measuring divergence between extant populations and calculating “evolutionary time” by summing the branch lengths separating species using time-calibrated phylogenies. What does this look like.
20Microevolutionary data yellow is allochronic and green is synchronic.
21Fossil dataWe’ve expanded the use of the time-series from Estes and Arnold by including longer term trends, and not just autonomous divergence.
22Comparative data Mammals And now we’re adding the comparative data for mammals.
23Comparative dataBirdsFade the mammals, add birds.
25Mammalian, Avian and Squamate body size Comparative dataMammalian, Avian and Squamate body sizeAnd all together now.
26To give a sense of what log-scaled divergence means To give a sense of what log-scaled divergence means. This time-series tracks the evolution of Hyracotherium, the dawn horse, to modern horses. This is a 5-7 fold increase in linear body size dimensions.
27“The Evolutionary Blunderbuss” ™ Stevan J Arnold Comparative dataMammalian, Avian and Squamate body sizeAnd all together now.
28If we look at within groups, not only is the pattern consistent, but the timing as well.
41Why millions of years? What process accounts for this pattern?
42What pattern makes evolution look like Brownian motion over macroevolutionary time? Genetic drift (Lande 1976)?No, parameters do not workDrift of a local adaptive peak?No, or we could never measure selectionA statistical pattern in need of an interpretation
43What then, is randomly walking? “Niches”Not a real answerWe need more focus on interpretation and causation!
44Is there anything significant about 1-5 milllion years? Species life-spans for mammals are ~1-5 million years (Alroy et al. 2000; Liow et al. 2008)Marine inverts my (Foote, 2007)Taxon cycles are suggested to occur on the order of my(Ricklefs & Bermingham 2002)
49What role could demography & extinction play? Similar to models proposed by Eldredge et al. (2005)As fitness-related mean trait value deviates from θ, mean population fitness declines according to a Gaussian function.
50Population growth (Burger & Lynch 1995) Let the population grow at a multiplicative growth rate, Rt, such that:andwhen, otherwisewhere B is the maximum population growth rate and K is the carrying capacity.
51Constant population size: Divergence increases with time Addition of demography:Divergence decreases with timeG/(ω2 +1)=0.01 σ2ө = 2.25 λ = 1/250, Linear increase in average divergence expectedG/(ω2 +1)=0.01 σ2ө = 2.25 B=1.025λ = 1/250 N0 = 10000
59From adaptive landscapes & G-matrices to phylogenetic stochastic models Need to account for the scaling from populations speciese.g. Futuyma’s ephemeral divergence modelDemography, and the dynamics of species ranges over time (including speciation) can be factored in to explain shifts in species meansThe ultimate “cause” of evolutionary change is change in adaptive landscapes
60Key parametersBirth rate must be low such that populations grow slowly, even when optimalSelection and inheritance parameters must not be so strong as to prevent any divergence at all, or the model fails to explain rapid divergence over short time scales.