Biology · Species, Speciation & Macroevolution

Evolutionary rates, developmental constraints and contingency

A mean trait rises from 10 to 14 over four time units. The endpoint rate is one trait unit per time unit. That same average can describe four equal increments or a long plateau followed by one four-unit change. A rate summarises a particular comparison; it does not reveal every step or cause.

Evolutionary Processes & Adaptation supplies allele frequencies, selection, drift, gene flow and the distinction between observation and explanation. Heredity, Variation & Evolution introduces branching ancestry. Meiosis & Patterns of Inheritance helps with the chromosome-set model; Development, Differentiation & Stem Cells helps with constraints and historical context. The examples require proportions, weighted averages and reading simple trees; each rule is explained where it is used. Species definitions for sexual organisms are introduced with their limits rather than assumed to classify every form of life.

Back to the Biology map

By the end, you can

Specify an evolutionary rate's numerator and timescale, then distinguish patterns of change from developmental accessibility and historical dependence.

Ask what changes and what the denominator counts

An evolutionary rate might describe trait change per unit time, sequence substitutions per site per time, or lineage origins per unit lineage-time. These quantities have different numerators and units. A high trait-change rate does not automatically imply a high speciation rate, and a species-rich group need not have had the highest recent origin rate.

In the preceding complete history, four lineages persist for two million years, five for four, six for one, and five for three. Total lineage-time is 4×2+5×4+6×1+5×3=49 lineage-million-years. There are two splitting events and one extinction event.

For that fully observed constructed window, the event-count ratios are 2/49≈0.0408 splits and 1/49≈0.0204 extinctions per lineage-million-year. Their difference is 1/49≈0.0204. These summaries use observed exposure, unlike the raw net count change of 1/10=0.1 lineages per million years. They are not fitted natural rates or evidence that either process was constant over time.

Rates inferred from incomplete natural histories require models, sampling assumptions and suitable additional evidence. Work on time-calibrated living trees has shown limitations on uniquely reconstructing general past speciation and extinction histories without further constraints. State those constraints instead of interpreting one fitted curve as an automatically unique history.

Mean endpoint change can hide a different tempo

The evenly changing model has values 10,11,12,13,14 at times 0,1,2,3,4. Its four observed interval changes are each one unit. The concentrated-change model has 10,10,10,14,14: its interval changes are 0,0,4,0. Both have the same endpoint increase of four over four time units.

Stasis describes limited net change in a measured character over a stated interval, often with variation around a relatively stable value. It does not mean that every gene, behaviour or ecological interaction stopped evolving. Fossil studies compare trait trajectories against alternatives such as directional change and fluctuations; the supported pattern can depend on the trait and the temporal resolution.

Punctuated-equilibrium discussions concern long intervals of relative morphological stasis and geologically concentrated change associated with lineage branching. Geologically rapid does not necessarily mean one generation or a single large mutation. Our concentrated curve has no supplied split, so it illustrates tempo but cannot by itself establish that full branching-associated pattern.

A connecting line between two samples is a visual aid. It does not measure what happened between them. Gaps, uncertainty and replacement by another lineage must be investigated before converting an abrupt-looking sampled difference into a mechanism.

One endpoint rate can summarise different paths

Two sampled mean-trait trajectories share axes and endpoints while distributing their observed change across different intervals.

Read the diagram in words

At times 0, 1, 2, 3 and 4, the even series has values 10, 11, 12, 13 and 14, with changes 1, 1, 1 and 1. The concentrated series has 10, 10, 10, 14 and 14, with changes 0, 0, 4 and 0. Both endpoints rise four trait units over four time units. No branching event or observation between samples is assigned.

Even sampled change
Model time units	Mean trait in model units	Change from previous sample
0	10	—
1	11	1
2	12	1
3	13	1
4	14	1

Change concentrated in one sampled interval
Model time units	Mean trait in model units	Change from previous sample
0	10	—
1	10	0
2	10	0
3	14	4
4	14	0

Both endpoint rates: (14−10)/4 = 1; Lines connect samples; intervening change is not measured; No lineage split or causal mechanism supplied

Two constructed trajectories of a population's mean trait in model units. Equal sampling times are known; lines connect samples only for comparison and do not establish continuous change between them. No lineage split, genotype data, error bars or causal mechanism is supplied.

The available combinations can be structured

Development connects inherited changes to phenotypes through interacting processes. A variant may change several traits together, and some imagined combinations may not be produced by the available developmental routes. Such constraints influence which variations selection can act on; they are different from a deliberate design goal.

The toy rule uses one variable z. Length=1+2z and width=1+z, with z restricted to 0,1,2. The permitted phenotypes are (length,width)=(1,1),(3,2),(5,3). An imagined long narrow form (5,1) cannot be generated by changing z alone.

That restriction belongs to the stated mapping. Another regulatory change or developmental pathway could change the relationship. Observing a few correlated traits in real organisms would not prove that every alternative is impossible forever. To explain a constraint, test the mechanisms and available variation rather than equating 'not yet observed' with 'cannot exist'.

One rule supplies only three trait combinations

A length–width plot shows three accessible combinations and one imagined combination absent from the supplied z-only mapping.

Read the diagram in words

z=0 gives length 1, width 1; z=1 gives 3,2; z=2 gives 5,3. The imagined target 5,1 cannot be generated with those permitted z values and that single mapping. A different developmental rule could change the accessible combinations.

Produced by this rule
Assigned z	Length in model units	Width in model units
0	1	1
1	3	2
2	5	3

Imagined independent combination
Relationship to the supplied rule	Length in model units	Width in model units
Not produced by changing z alone	5	1

Length = 1 + 2z; width = 1 + z; Allowed z values: 0, 1 and 2; No fitness or universal impossibility assigned

An invented developmental rule with one variable z taking values 0, 1 and 2. Length = 1 + 2z and width = 1 + z in model units. All allowed phenotypes follow this rule; no viability, fitness or real organism is assigned. Omitted regulatory changes could change the rule.

An earlier background can change a later variant's effect

Historical contingency means that an outcome depends on earlier events and conditions. In the assigned two-marker case, F has fitness 0.70 when earlier background variant M is absent, compared with 1.00 without F. With M present, F has fitness 1.40, compared with 1.00 without F. The same later variant reverses its relative effect across backgrounds.

Starting with 100 individuals of each F-bearing background gives expected outputs of 70 and 140. These are assigned expectations in a fixed environment, not observed counts. The comparison illustrates genetic-background dependence without saying M arose in preparation for F.

Some outcomes can recur in similar contexts while other details depend strongly on history. Repeated habitat-associated forms in the Anolis research and replay comparisons in the preceding evolution chapter provide different ways to test that relationship. Evidence about one contingency does not make all evolution unpredictable, and a repeated pattern does not eliminate every historical influence.

The same variant has a different effect on two backgrounds

A two-by-two table and paired expectation bars compare F-bearing and no-F states separately for each earlier M background.

Read the diagram in words

Without F, expected offspring per starter are 1.00 in both M backgrounds. With F they are 0.70 when M=0 and 1.40 when M=1. For 100 starters of each state, outputs are 100 and 70 without/with F in M=0, and 100 and 140 in M=1. The same F variant reverses its relative effect across the assigned backgrounds.

The same variant has a different effect on two backgrounds
Earlier M background	F state	Expected offspring per starter	100 starters for each state	Expected offspring, not measured counts
M = 0	Without F	1	100	100
M = 1	Without F	1	100	100
M = 0	With F	0.7	100	70
M = 1	With F	1.4	100	140

F relative to its no-F reference
Earlier M background	F relative to its no-F reference	Effect in this background
M = 0	0.7 / 1 = 0.7	F is lower than its no-F reference when M is absent
M = 1	1.4 / 1 = 1.4	F is higher than its no-F reference when M is present

Earlier history changes effects; no future-directed mutation

An invented haploid two-marker model. M describes an earlier background variant; F is a newly introduced variant. Fitness is expected counted offspring per starting individual in one fixed environment. Values are assigned, not measured; no future-directed mutation or inevitable pathway is assumed.

Worked example

Compare a summary with the observations it compresses

For both trait series, (14−10)/(4−0)=1 model trait unit per time unit. With only the first and last values, the two histories would be indistinguishable in this comparison. Intermediate samples show evenly distributed change versus one concentrated interval.

For the F comparison, divide by the no-F reference of 1.00 in each background. F's relative values remain 0.70 and 1.40: lower in the first, higher in the second. Neither the trait endpoint rate nor this fitness comparison tells us how many new species formed.

Check your understanding

Why is 2/49 from the lineage ledger not interchangeable with 1/10?
Show reasoning
The first counts splits per lineage-million-year; the second counts net added lineages per million years. They use different numerators and denominators, so they answer different questions.
Do the identical endpoint rates prove the two trait series changed in the same way?
Show reasoning
No. Both average one unit per time unit, but their intermediate observed changes are 1,1,1,1 and 0,0,4,0. A summary rate can conceal a different tempo.
Why does absence of the (5,1) phenotype in the developmental toy not establish its universal impossibility?
Show reasoning
It cannot be produced by the supplied z-only mapping. Omitted regulatory or developmental changes could alter that mapping; the restriction is conditional, not a proof about all organisms or future variants.

What to carry forward

Define the rate and comparison, inspect the path, and keep developmental and historical conditions attached to the explanation.

References

Resources supporting this lesson’s explanations and examples.