Phylogeny

Read ancestry from nodes, not from the order of the tips

A rooted phylogenetic tree has a direction from an ancestral lineage toward descendants. Each branch represents a lineage, each tip identifies a sampled taxon or lineage, and an internal node represents a hypothesized common ancestor at a divergence. Two lineages that meet at the same immediate node are sister groups.

To compare A with B, trace both branches backward until they first meet. Then do the same for A and C. If the A–B meeting point is more recent than the A–C meeting point, A is more closely related to B. Horizontal distance between printed labels does not answer the question, and counting every node along two paths can fail when one part of the tree contains more sampled lineages.

Branches may be rotated around a node without changing ancestry. Thus (((A,B),(C,D))) and (((B,A),(D,C))) have the same topology. Changing which names appear on the left or top is like turning a mobile; it does not reconnect its pieces.

A clade includes an ancestor and all of its descendants

A clade, or monophyletic group, begins at one node and contains every lineage descending from that node. On a printed tree, a reliable test is to imagine cutting one branch: everything that falls away together forms a clade. Selecting descendants from one node while leaving out another descendant does not produce a complete clade.

A polytomy is a node drawn with more than two descendant branches. It commonly represents unresolved branching order; it does not necessarily assert that all lineages split at exactly the same instant. New evidence can resolve a polytomy, and conflicting evidence can replace one fully resolved tree with another.

No living tip is automatically the ancestor of another living tip. A lineage drawn near the root is not “less evolved,” and a lineage on the far right is not a goal toward which the others progressed. All living lineages have histories extending from their shared ancestors to the present.

Cladograms and scaled trees do not encode the same information

“Phylogenetic tree” is a broad term, so never infer time merely because one line is longer. In an unscaled cladogram, a long branch may have been stretched only to fit a label. A time estimate requires a clock model or another calibration, not a ruler placed on an arbitrary drawing.

Build a tree from shared derived characters

A useful character must compare homologous features. An analogous similarity produced independently by convergent evolution can mislead the analysis. An outgroup is a related lineage outside the focal ingroup; comparison with it helps root the tree and infer which state may be ancestral, provided the chosen outgroup and evolutionary assumptions are reasonable.

The following invented matrix uses 0 for the reference state observed in outgroup O and 1 for a derived state. Characters (c_1) through (c_5) are abstract so the exercise tests tree logic rather than memorized anatomy.

Character (c_1) supports an ingroup containing A–D. Character (c_2) is a shared derived character supporting the A–B clade, while (c_3) supports the C–D clade. Characters (c_4) and (c_5) distinguish B and D but do not, by themselves, establish a sister pair.

Molecular evidence adds many comparable characters

DNA and protein sequences supply hundreds or thousands of positions that can be compared among lineages. Before counting differences, researchers must identify homologous sequences and align positions that descend from corresponding ancestral positions. A raw similarity score can be misleading if unrelated regions are aligned or if insertions and deletions are handled poorly.

Molecular data often resolve relationships that morphology alone cannot, but a gene tree is not automatically identical to a species tree. Horizontal gene transfer, hybridization, gene duplication, gene loss, and incomplete lineage sorting can give individual genes different histories. Confidence grows when independently informative loci, fossils, and morphology support the same branching pattern.

A tree should therefore be read as a model supported to a stated degree, not as a photograph of the past. New sequences or a better substitution model can revise the topology without making the earlier evidence unscientific.

A molecular clock needs both a rate and a calibration

If substitutions accumulate at an approximately stable rate, sequence divergence can help estimate the time since two lineages separated. Let (d) be corrected pairwise divergence in substitutions per site, (r) the rate in substitutions per site per million years for each lineage, and (t) the time since divergence in million years. With equal rates on the two descendant branches,

For an invented example, take (d=0.060) substitutions per site and (r=0.0020) substitutions per site per million years per lineage:

The factor of 2 appears because changes accumulate along both descendant lineages. If a source defines its rate as the combined pairwise divergence rate, the formula would be written differently, so the unit and rate definition must be read first.

This estimate is conditional. It assumes a suitable calibration, approximately clock-like rates, homologous sequence, and correction for substitutions that occurred more than once at the same site. Fossils can provide minimum or maximum constraints, but a fossil's age is not automatically the exact date of the node.

Make only the inference the tree supports

The tree does not show which lineage is “best,” whether a trait is beneficial, or how much present-day species resemble their ancestors unless those data are separately encoded. It organizes hypotheses about descent; biological explanations still require the evidence used to build and test it.

Check your tree reasoning