Questions: Comparative Phylogenetic Methods for Evolutionary Analysis

Question 1

A biologist collects brain size and body size data from 200 mammal species and runs a standard regression, finding a strong positive correlation. A reviewer argues the analysis is statistically flawed. What is the most likely objection?

• A. The sample size of 200 is too large, inflating the apparent statistical significance
• B. Brain size and body size are not on comparable measurement scales for regression analysis
• C. Species share evolutionary history through common descent, so they are not statistically independent observations — including many closely related species inflates the apparent sample size without adding independent evidence
• D. A regression cannot be used for biological data; only ANOVA is appropriate for cross-species comparisons

Answer: (C)

Regression assumes data points are independent observations. Species are not independent: closely related species share a common ancestor who may have already had the trait, so including 20 primate species is not like having 20 independent tests — it largely re-measures the same evolutionary event multiple times. Phylogenetically independent contrasts solve this by measuring trait changes at branching points, each of which represents one genuinely independent evolutionary divergence.

Question 2

Two researchers test whether species that evolve bright coloration also evolve toxicity. Researcher A plots 80 species and finds a significant correlation. Researcher B transforms the data into phylogenetically independent contrasts and re-runs the analysis. Whose approach is more statistically valid?

• A. Researcher A's, because more raw data points provide greater statistical power and more accurate estimates
• B. Researcher B's, because PICs measure genuinely independent evolutionary events rather than species that may share ancestral traits through common descent
• C. Both are equally valid because large sample sizes automatically correct for phylogenetic non-independence
• D. Neither, because hypotheses about correlated evolution cannot be tested with statistical methods

Answer: (B)

PICs transform species comparisons into contrasts measured at each node of the phylogeny — at each branching point, the contrast between sister lineages represents one genuinely independent evolutionary event. Researcher A's raw species plot may have pseudo-replication: if all toxic-and-bright species are closely related, the apparent pattern may reflect a single ancestor that evolved both traits, not many independent co-evolutions of brightness and toxicity.

Question 3

Including 20 primate species in a cross-species dataset provides 20 independent data points for testing an evolutionary hypothesis, equivalent in statistical value to 20 data points from distantly unrelated species.

• T. True
• F. False

Answer: False

Primates share a recent common ancestor, so much of their similarity reflects inherited ancestral traits rather than 20 independent evolutionary outcomes. Including them inflates apparent sample size without adding proportional statistical independence. PICs correct for this by measuring contrasts at branching points — each branching event represents one independent evolutionary comparison, regardless of how many species subsequently descended from those lineages.

Question 4

Phylogenetic signal quantifies how much of the variation in a trait is predicted by evolutionary history; a trait with high phylogenetic signal is one where close relatives tend to resemble each other more than expected by chance.

• T. True
• F. False

Answer: True

High phylogenetic signal (reflected in high Pagel's λ or Blomberg's K) means the trait distribution across the tree is well-predicted by phylogenetic relatedness — close relatives are more similar than distant relatives, beyond what would be expected if the trait evolved independently of ancestry. This suggests evolutionary conservatism, heritable constraints, or common ancestral origins for the trait, rather than trait values being driven primarily by current ecology independently of ancestry.

Question 5

Why do two closely related primate species provide less independent evidence for an evolutionary hypothesis than two distantly related mammalian lineages? What do phylogenetically independent contrasts do to address this?

Model answer: Closely related species share a recent common ancestor, so their trait values are not statistically independent — they largely reflect inherited ancestral characteristics rather than independent evolutionary outcomes. Two primate species that both have large brains may both have inherited that trait from a single ancestor, not evolved it independently. Phylogenetically independent contrasts address this by calculating trait differences between sister lineages at each branching node of the phylogeny. Each contrast at a node represents one genuinely independent evolutionary divergence event — the moment when two lineages split and began accumulating independent changes. Running correlations on contrasts rather than raw species values ensures that each data point reflects an independent evolutionary test of the hypothesis.

Felsenstein's 1985 introduction of PICs was transformative because it revealed that many classic cross-species comparisons were statistically unsound. Even well-known patterns had to be re-tested. The principle applies broadly: any time you want to test whether two traits co-evolve across species, you must account for the fact that species similarity can reflect shared ancestry rather than independent evolution.