The classic way to categorize speciation barriers: there are two ways to stop reproduction. Stop the wedding, or kill the baby. Stop the wedding — anything that prevents copulation: physical separation, behavioral mismatch, anatomical mismatch. Kill the baby — anything that lets the zygote form but kills it: blastula failure, gastrula failure, dies as a fetus, dies as a newborn before reproducing. Pre-zygotic and post-zygotic, in the fancy version. Just remember: stop the wedding, kill the baby.
— 202_lec26_03
A — Two populations diverging neutrally
The big thing with species is to what extent these barriers just happen versus to what extent they're actively selected for. This bothers people at first: the vast majority of speciation has absolutely nothing to do with selection. We're not selecting for those differences. Drift is random, mutations are random, and differences accumulate over time. If two things are separated for long enough, when they come back together they often just don't work.
— 202_lec26_04
Predict first. Two lineages drift apart, each fixing changes that work fine at home. Double the time apart, and the number of combinations that fail together will:
Each lineage fixes changes over time. A change from one lineage has never been tested beside a change from the other, and some of those pairings do not work. Count the pairs, not the changes.
at 40 changes each: —
at 80 changes each: —
doubling the time multiplied the failures by: —
at 80 changes each: —
doubling the time multiplied the failures by: —
B — Count the pairs, watch the snowball
As soon as I invert a region, I've created a barrier — not between individuals, but between chunks of a chromosome. There can no longer be gene flow between different chromosome chunks. Inversions are the seed of speciation.
— 202_lec25_06
TODO: incompatibility-snowball sim. log-log plot of incompatibilities vs divergence time; slope of 2. SPEC: Each substitution on lineage A creates potential incompatibility with each substitution already accumulated on lineage B. The number of incompatibilities scales as n² where n is the number of substitutions per lineage. Hybrid fitness drops faster than linear.
C — Reinforcement closes the gap
Reinforcement: the existence of a post-zygotic barrier (subfertile hybrids) creates positive selection for a pre-zygotic barrier (mate preference). You're reinforcing the barriers with each other — making a stronger combined barrier, more quickly independent. But it's a race between selection and recombination. You need subfertile heterozygotes that aren't too bad, and you need a preference, and you need the preference and the reason for subfertility to be linked, so recombination can't unscramble it.
— 202_lec26_05
TODO: reinforcement sim. Selection-vs-recombination race; show when reinforcement succeeds and when it fails. SPEC: Once post-zygotic incompatibility is in place, selection favors pre-zygotic isolation — because hybrids are wasted reproductive effort. Specify the test: when does reinforcement happen, and when does recombination break the link between the trait that causes the incompatibility and the trait that signals mate choice?
D — Hybrid fitness across divergence — real data
I will either have long-tailed squirrels or I will have short-tailed squirrels. That sounds like a boring world. I want a world where I have long-tailed squirrels and short-tailed squirrels. For that world to be, my medium-tailed squirrels can't breed. I need to stop making them.
— 202_lec27_03
TODO: cross-taxon snowball plot. Fit power-law exponent to each dataset; show clustering around 2. Non-trivial code mod: fit the exponent as a free parameter and ask whether 2 is in the resampled range. SPEC: Compare hybrid fitness as a function of divergence time across taxa. Drosophila, sunflowers, sticklebacks. The snowball pattern (t² scaling) is detectable in real data.