Species-Level Diversity: More Than Population-Level Processes?
Macroevolution · PB 495/595 Plant Evolutionary Biology
29 October 2026
Species are born and die
- Species are separately evolving lineages (de Queiroz 2007).
- A lineage has a beginning (speciation), an end (extinction), and traits that its daughter species inherit.
- If some species leave more daughters, or last longer, this reflects the concepts of how differential survival and reproduction changes a population (Vrba and Gould 1986; Jablonski 2008).
How many plant species?
- Angiosperms are about 90% of land-plant species.
- They are relatively young. The crown of angiosperms is much younger than gymnosperms (Magallón et al. 2015).
- Is this surprising?
Species per angiosperm order
Five of the 65 orders hold almost half of all angiosperm species.
Species per family
- The five largest families hold a third of all angiosperm species: Asteraceae, Orchidaceae, Fabaceae, Rubiaceae and Poaceae.
- The median family has 52 species, and 128 families have 10 or fewer.
- Early-diverging lineages are often small, like Amborella (1) and Acorus (5).
Species per genus
- Most genera are tiny, but a few are very large (Willis and Yule 1922).
- About 28% of Angiosperm genera have a single species, but the bulk of diversity comes from some of the larger families we have discussed (Frodin 2004).
- What could explain this? Time, rates of speciation that differ among lineages, key innovations, ecological opportunity.
The Yule process
- Every lineage splits at rate \(\lambda\), independently of the others. Nothing goes extinct.
- With \(k\) lineages alive, a split happens at total rate \(k\lambda\), so the wait until the next one is
\[ T_k \sim \text{Exponential}(k\lambda), \qquad E[T_k] = \frac{1}{k\lambda}. \]
- Each split adds a lineage, so the waits get shorter as the tree grows.
Yule growth is exponential
Each lineage adds new lineages at rate \(\lambda\), so on average
\[ \frac{dN}{dt} = \lambda N \quad\Longrightarrow\quad E[N(t)] = N_0\, e^{\lambda t}. \]
- On a log axis that’s a straight line with slope \(\lambda\).
- Single runs scatter a lot around the average, mostly early on, while only a few lineages exist.
Yule’s explanation of that heavy tail
- Start a genus from one species. After time \(t\) its size is geometric:
\[ P(N(t) = n) = e^{-\lambda t}\left(1 - e^{-\lambda t}\right)^{n-1}. \]
- Young genera are mostly small and old ones can be huge. New genera keep appearing, so most genera are young (Yule 1925).
- Mix genera of all ages and you get the hollow curve. The simulation looks a lot like the real data.
Exceptional clades: diversity envelopes
- Pick a background rate, such as the angiosperm average \(\hat r = 0.085\).
- For each clade age, work out the richness you’d expect at that rate. The bands hold 95% of outcomes, for low and high turnover (Magallón and Sanderson 2001).
- Clades above the band are unusually rich, and clades below it are unusually poor.
- Can you name a clade that is more or less than the average?
The paradox of stasis
- Evolution measured over a few generations can be fast (Hendry and Kinnison 1999), but measured rates drop as the time interval gets longer (Gingerich 1983).
- In real data, divergence stays bounded for about a million years and only then builds up. Populations tracking an optimum that occasionally jumps fits that pattern (Uyeda et al. 2011; Estes and Arnold 2007).
Example: the million-year wait
Uyeda et al. (2011) put together more than 8,000 body-size measurements, from field studies to fossils to comparisons among living species.
Fig. 1 from Uyeda et al. (2011)
Example: why is self-incompatibility still around?
Within populations, selfing should win
- Selfing reduces genetic diversity but gives reproductive assurance.
- Self-incompatibility (SI) is lost again and again, and the loss is close to irreversible (Igić, Lande, and Kohn 2008).
Across lineages, SI hangs on
- In Solanaceae, SI lineages diversify faster than self-compatible (SC) lineages (Goldberg et al. 2010).
Example: why is self-incompatibility still around?
The Binary State-dependent Speciation and Extinction (BiSSE) model. A preview of later activities.
Fig. 1 from Goldberg et al. (2010)
Speciation and extinction on a tree
Lineages split (blue dots) and some die out (crosses). We only sample living species, so the tree we estimate is the reconstructed tree on the right. Extinct lineages leave no branches in it.
Birth–death, previewed
Model a clade as lineages that speciate at rate \(\lambda\) and go extinct at rate \(\mu\):
\[ E[N(t)] = N_0\, e^{(\lambda-\mu)t}, \qquad r = \lambda - \mu, \qquad \epsilon = \mu/\lambda . \]
Lineage-through-time plots
- Slide a line from the root to the present and count the lineages it crosses. Each split adds one.
- Plot that count on a log axis. Under a constant-rate Yule process you’d expect a roughly straight line with slope \(\lambda\).
- An LTT is a quick look at the tempo of a tree before fitting any model.
Reading LTT shapes
- With extinction, the curve bends up near the present.
- A slowdown can mean diversity-dependence.