Branching relationships matter; tip order, visual proximity, and a ladder of progress do not.
By the end: Find common ancestors, interpret clades, and distinguish species history from individual gene histories.
A rooted evolutionary tree represents hypotheses about descent. Tips represent sampled lineages; an internal node represents a common ancestral lineage or population. Two tips are closer relatives when their most recent common ancestor is more recent than the ancestor they share with another tip.
A clade contains an ancestor and all its descendants. Rotating branches around an internal node changes the drawing order but not the ancestry relationships. The two trees in the experiment let you compare a rotation with an actual change in grouping. Branch lengths here are schematic, not elapsed time or amount of genetic change.
Begin at a tip and trace its branch toward the root. Each encountered node joins that lineage to additional descendants. In the grouping of A with B and C with D, the first shared node reached from A and B has only A and B as sampled descendants. The root joins all four. This gives a precise meaning to closer relationship. A and B share a more recent common ancestral lineage than either shares with C. The drawing does not say that A resembles B more in every measurable trait.
A rooted tree records nested descendant sets. For this example those sets include A with B, C with D, and all four together. Swapping A and B on the page preserves every set. Swapping entire pairs also preserves them. In contrast, grouping A with C replaces an internal descendant set and changes the hypothesis. This is why rotating a node is harmless but moving a tip to another branch is not. Check the membership of the sets instead of comparing the horizontal or vertical distances between letters.
A clade must include all descendants of its chosen ancestor, including descendants omitted from a simplified drawing. If a diagram shows only A, B, C, and D, circling A and B can identify a clade relative to that sample. A fuller tree might place additional sampled lineages inside the same group. Excluding one of those descendants would no longer represent that complete clade. Sampling changes which branches are visible, not which actual organisms descended from an ancestor. A diagram's scope belongs in the interpretation of any highlighted group.
The root adds direction to ancestry. An unrooted tree shows connections among sampled lineages without specifying where their common ancestry begins. Researchers may use an outgroup, a lineage inferred from other evidence to lie outside the focal group, to help place a root. The outgroup is not the ancestor of the other tips. It has its own history after divergence. Uncertain root placement can leave the direction of character change or order of ancestral splits unresolved even when several connections are well supported.
Branch length has meaning only when the figure declares a scale. A schematic tree can stretch a branch to fit a label without changing its claim. A time-scaled tree instead uses distance along its time axis to represent elapsed time. A sequence-change tree may scale branch lengths by estimated substitutions per site. Equal lengths in that last tree do not necessarily imply equal durations because evolutionary rates can differ. Never extract an age or a rate from the geometry of the lesson's unscaled four-tip drawings.
| Tree type | Length can represent | Cannot infer without more information |
|---|---|---|
| Schematic topology | Drawing space only | Divergence dates or change amounts |
| Time-scaled tree | Elapsed time along the declared axis | Substitution rate without sequence change |
| Sequence-change tree | Estimated substitutions per site | Elapsed time without a rate model |
Use the existing tree presets to test this reading method. Before choosing the rotation preset, list the two sister pairs and the sampled descendants of the root. They remain the same afterward, even though the tip order changes. Then choose the different-groups preset and find the common ancestor of A and B again. It is now the root rather than their own pair node. The widget displays supplied hypotheses and marks this ancestor. It does not calculate those hypotheses from DNA, date any node, or estimate confidence from a dataset.
Humans did not descend from today’s chimpanzees; the lineages share ancestral populations. Living species have their own histories since divergence. A tip drawn near the root or on the left is not automatically more primitive or less evolved. Some traits can retain ancestral forms while other traits change.
Similarity can reflect shared ancestry or independent evolution under similar pressures. Homologous features are compared as inherited from a common ancestor; analogous functional similarities need not share that specific origin. A derived character is a change relative to an inferred ancestral state. A shared derived character can support grouping its bearers when repeated origin or loss is less plausible. Scientists compare many characters and explicit evolutionary models to infer trees rather than rank organisms by one resemblance.
A living lineage drawn near the root has not stopped evolving. Its branch represents descendants continuing since their separation from the other lineages, whether or not the drawing shows intermediate events. One trait can retain an ancestral form while another changes. Calling the organism ancestral confuses one character's state with the whole lineage's history. Similarly, an earlier branching position does not measure complexity or success. Branching concerns ancestry, not a ranking or a sequence each organism must pass through.
To infer a character's history, first define comparable states. Imagine an outgroup O and four focal lineages. O, C, and D have state zero, while A and B have state one. If A and B are sisters, one change on their shared ancestral branch can explain the pattern. This makes state one a possible shared derived character for that group. Derived means different from the inferred ancestral state. It does not mean better, more complicated, or more recently evolved in every part of the organism.
Compare a competing tree that groups A with C and B with D. With the root assigned state zero, explaining state one in A and B now requires at least two changes under a simple binary model. These could be separate gains, or a gain followed by losses in a suitable reconstruction. Counting the minimum changes is a method called parsimony. It is useful for seeing what a character pattern favors, but the minimum count is not a guarantee of historical truth. Repeated changes can genuinely happen, and characters can differ in their rates.
Homology is a hypothesis of inherited correspondence, not a claim that present functions are identical. Two structures can descend from the same ancestral structure and later perform different tasks. Conversely, similar performance can arise independently in lineages facing similar conditions. That similarity is convergence when the relevant state has separate origins. A structure can be homologous at one level of comparison while a particular specialized shape is convergent. Define exactly which feature is being compared before deciding whether shared ancestry or independent change best explains it.
Current usefulness does not date the origin of a trait. Suppose a surface that originally had another function now improves movement through water. Demonstrating that performance today supports a present functional claim. To argue that selection originally favored the surface for swimming, we also need historical evidence about when the surface appeared, its earlier contexts, and plausible alternatives. Fossils and comparisons can constrain that history without preserving every event. A tree supports questions about the sequence of changes; it does not turn a plausible adaptive account into an observed ancestral event.
| Question | Relevant evidence | Unresolved by that evidence alone |
|---|---|---|
| Does a feature help now? | Performance and reproductive consequences today | Its original function |
| Was the feature inherited? | Corresponding structures and sequence history | Why selection favored it |
| Was it selected for this use? | Historical context plus inherited fitness effects | Every unpreserved ancestral event |
| Could another origin explain it? | Earlier functions, by-products, repeated origins | A unique story without comparisons |
For guided exploration, map the binary states on both supplied paper trees and mark each proposed gain or loss. Then reverse the proposed ancestral state and recount. Notice which conclusions depend on rooting and which depend only on the observed tip states. Finally add a second character with the opposite pattern. The evidence now conflicts, so neither character alone settles the tree. More comparable data and explicit assumptions about change are needed. Do not resolve that conflict by choosing the more useful-looking trait or declaring one living lineage less evolved.
A and B are sister lineages in both trees.
| Tree grouping | A and B sisters? | Effect of rotation |
|---|---|---|
| ((A,B),(C,D)) | Yes | None on ancestry |
| ((B,A),(C,D)) | Yes | Same grouping |
| ((A,C),(B,D)) | No | Different grouping |
A molecular tree begins with observations, not with the finished branching diagram. Researchers compare sequences thought to be corresponding inherited regions and align positions so that each column represents a proposed correspondence. An insertion or deletion can make alignment uncertain. Comparing different duplicated genes as though they were the same inherited copy can also mislead. These decisions occur before calculating a tree. The lesson's tree lab skips them deliberately because its purpose is reading relationships, not reconstructing them from molecular observations or evaluating alignment software.
Consider a constructed alignment containing six variable positions. At four positions, A and B share one base while C and D share another. At one position, A groups with C, and at one, A groups with D. Under an equal-cost model that permits one change per supporting split, the AB with CD grouping requires four single changes and two double changes, totaling eight. Either alternative requires eleven. This small calculation shows how multiple characters can favor one hypothesis while some disagree. The positions are invented, not measurements from actual species.
| Site pattern | Sites | AB | CD cost | AC | BD cost | AD | BC cost |
|---|---|---|---|---|
| A=B, C=D; groups differ | 4 | 4 × 1 = 4 | 4 × 2 = 8 | 4 × 2 = 8 |
| A=C, B=D; groups differ | 1 | 2 | 1 | 2 |
| A=D, B=C; groups differ | 1 | 2 | 2 | 1 |
| All six sites | 6 | 8 | 11 | 11 |
| Remove one AB site | 5 | 7 | 9 | 9 |
| Remove three AB sites | 3 | 5 | 5 | 5 |
The calculation assumes that all six sites are comparable and that extra changes have the same cost. Real sequence inference can use models that allow unequal substitution rates, repeated substitutions, and differing rates across positions. A site observed as the same base in two tips may have changed and changed back along one branch. Observed differences therefore do not count every historical substitution directly. A method can prefer a tree because it explains the data well under its model, while another reasonable model or alignment yields a different result.
Sampling uncertainty remains even with a specified method. A short sequence can omit informative changes by chance. One common support procedure resamples alignment columns with replacement and reconstructs trees from those resampled datasets. A split found repeatedly is stable under that resampling procedure. Its support percentage is not automatically the probability that the split is true, and resampling cannot correct a systematically wrong alignment or model. Strong support for one split also does not imply strong support for every other split or for the root's placement.
Independence needs a biological interpretation. Adjacent sites in a segment often travel together through reproduction, so thousands of bases can describe one local genealogy rather than thousands of independent gene histories. More sequence can clarify that genealogy while leaving disagreement with another region unresolved. Comparing separated genomic regions can test whether a branching pattern is widespread, but recombination and exchange still matter. Distinguish confidence in the inferred history of one segment from confidence that the same tree describes the sequence of population splits across the whole genome.
| Evidence or procedure | What it addresses | What it does not guarantee |
|---|---|---|
| More sites in one segment | Resolution of a local genealogy | Independent population histories |
| Resample alignment columns | Stability to column sampling | Probability the split is true |
| Compare separate regions | Agreement among local histories | No recombination or exchange |
| Try alternative models | Sensitivity to assumptions | That one tested model is correct |
Use the six-site comparison as a sensitivity exercise on paper. Remove one AB-supporting site, then recompute the costs. The favored grouping costs seven and the alternatives nine, so the preference narrows. Remove three such sites instead and all three groupings cost five. The example shows why an answer can depend on limited sampling without making all trees equally plausible. Ask what additional observations could distinguish the alternatives and whether they represent independent histories. No preset in the existing lab performs this inference or reports a statistical support value.
Fossils, anatomy, and molecular sequences provide evidence for history. Sequence comparisons require decisions about which positions correspond and how changes occur. Different models or limited data can leave branching order uncertain. Support for one branch is not certainty about every feature of a tree.
A gene tree can differ from a species tree because ancestral variants persist through splits, because genes move between lineages, or because of reconstruction error. Horizontal transfer can make a network more informative than one strictly branching history for some questions. Common ancestry is supported by converging evidence; a single diagram is not the whole argument.
A species tree describes relationships among populations or lineages, while a gene genealogy follows particular inherited copies. A diploid individual's two autosomal copies can have different recent ancestral paths. Looking backward, two sampled copies join when they descend from the same ancestral copy. That joining event can precede the split between their populations. Populations can separate while still carrying several variants inherited from an older population. A species split need not make gene copies join simultaneously.
Imagine that the population ancestral to A, B, and C contained two old variants, x and y. The species history first separates C from the ancestor of A and B, then separates A and B. If A retains x, B retains y, and C retains x, a sampled segment can group A with C. The persistence and sorting of ancestral variation is called incomplete lineage sorting. This history requires no later transfer. It requires tracing inherited copies, rather than assuming every sequence difference arose after the species carrying it became separate.
In a simplified neutral model, a rapidly successive pair of population splits leaves less opportunity for copies in the A and B ancestral population to join before reaching the older common population. Larger ancestral effective population size can also prolong the persistence of separate ancestral copies. These tendencies do not date the cartoon's splits. A conflicting gene tree could instead reflect exchange or inference error. Identifying incomplete lineage sorting in a real case requires comparisons among histories and models rather than labeling every conflict with the same explanation.
Introgression creates another route to disagreement. If an individual from C contributes descendants to A through hybridization and subsequent breeding, an inherited segment in A can have a more recent connection to C than the broader species history suggests. Recombination can leave that connection in only part of the genome. Horizontal transfer can produce another kind of cross-lineage connection without the same sexual ancestry. A network can show such added routes more directly than a strictly branching tree. Evidence must identify the transferred material and uncertainty.
Compare the three hypotheses before accepting a transfer story. Ancestral variation predicts discordance without a later movement event. Introgression predicts a history involving reproductive exchange and potentially localized ancestry patterns. Reconstruction error predicts that limited data or inappropriate assumptions produced an incorrect gene tree. Additional loci, population sampling, and models can discriminate among these alternatives without guaranteeing certainty. Shared sequence alone does not name the route. Likewise, a gene arriving from elsewhere does not mean that the recipient's entire genome or every ancestor followed that route.
| Candidate explanation | Possible route to the pattern | Evidence to compare |
|---|---|---|
| Ancestral variation | Old copies persist through population splits | Distribution of genealogies across regions |
| Introgression | Hybridization and later reproduction transfer a segment | Local ancestry and contact history |
| Horizontal transfer | DNA moves outside ordinary sexual ancestry | Sequence context and plausible transfer history |
| Reconstruction error | Alignment, model, or sampling gives a wrong tree | More data and sensitivity checks |
Finish by reading a hypothetical result with the claims separated. Suppose most sampled regions group A with B, while one well-supported region groups A with C. The observations support disagreement among inferred local histories. They do not alone prove that the species tree is wrong, that C was a direct ancestor of A, or that the shared sequence was adaptive. State the candidate routes and the evidence still needed. This preserves the distinction between an inherited history, a mechanism of exchange, and a functional story about why a variant later increased in a particular environment.
In the grouping ((A,B),(C,D)), swap the drawing positions of C and D. Which sister pair changes?
A node can rotate while keeping exactly the same descendants.
| Drawing | Sister pairs |
|---|---|
| ((A,B),(C,D)) | A/B and C/D |
| ((A,B),(D,C)) | A/B and D/C |
Neither pair changes. A and B remain sisters, as do C and D. The ordering of terminal labels is a drawing choice. Only changing which lineages share internal ancestors changes this topology.
Two distantly related lineages share a streamlined shape. Must they be sisters?
Ask whether similar function can favor independent changes.
| Possible explanation | Evidence to compare |
|---|---|
| Shared recent ancestry | Other inherited shared features |
| Convergent form | Independent origins in broader trees |
| Uncertain reconstruction | More data and model checks |
No. Similar environments can favor similar shapes independently, a form of convergence. Evaluate ancestry using many informative features and models that allow repeated changes. A functional resemblance alone does not override the broader evidence for their relationships.
The explanations and exercises are complete without these optional links. The references cover inheritance, population mechanisms, and evidence for ancestry. Source annotations distinguish textbook explanations, secondary history, and a primary paper whose bibliographic identity alone was checked. The examples are constructed teaching models, not empirical estimates or professional decision advice.
Rooted trees, common ancestry, and reading branching relationships. Our four-tip trees are hypothetical and not time-scaled.
Descent with modification, selection, evidence, and common misconceptions. Numerical examples here are invented, not field measurements.