The Evidence for Evolution

Evolution by descent with modification predicts that all living organisms share common ancestors, more or less remote depending on how related the organisms are. This single claim generates an enormous number of independent, falsifiable predictions. If two species share a recent common ancestor, they should share derived similarities (called synapomorphies) inherited from that ancestor — in their bones, their embryonic development, their biogeographic history, and their DNA. Crucially, these predictions were testable using methods unavailable to Darwin, most importantly molecular sequencing, which did not exist until roughly a century after On the Origin of Species (1859). A key conceptual distinction runs through all the evidence: homology versus analogy. Homologous structures are similar because of shared ancestry — the forelimb bones of a bat, whale, cat, and human are built from the same ancestral bauplan (humerus–radius/ulna–carpals–digits) even though they now function as a wing, a flipper, a paw, and a hand. Analogous structures, by contrast, are similar because of convergent adaptation to similar selective pressures, not shared ancestry — the wings of insects, birds, and bats evolved independently and share no common winged ancestor. Distinguishing homology from analogy is the central analytical skill in reading comparative evidence, and it is exactly what phylogenetic methods (see Phylogenetic Trees) are built to do systematically. A second organizing concept is that evidence for evolution is best understood as a set of convergent, independently falsifiable tests. The fossil record could in principle have shown mammal fossils below fish fossils, or fully formed elephants appearing with no older relatives whatsoever — it doesn't. Biogeography could have shown that isolated oceanic islands host the full range of terrestrial vertebrate classes seen on continents — it doesn't (islands like the Galápagos and Hawaii famously lack native land mammals, amphibians, and reptiles other than what could plausibly raft or fly there). Molecular sequences could have shown humans more similar to birds than to chimpanzees at the DNA level — they don't. Each of these is a place evolutionary theory could have been falsified and was not.

Step 1 — Stratigraphy establishes relative and absolute time. Sedimentary rock accumulates in layers (strata), with older layers beneath younger ones (the geological law of superposition). Radiometric dating — measuring the decay of isotopes such as potassium-40 to argon-40 (half-life ≈ 1.25 billion years) or carbon-14 (half-life ≈ 5,730 years, useful only for material under ~50,000 years old) — assigns absolute ages to rock layers and the fossils embedded in them. This gives paleontologists an independently derived timeline against which anatomical claims can be checked.

Step 2 — Fossilization captures anatomical snapshots through time. Fossils form when hard parts (bone, shell, teeth) are buried rapidly and mineralized before decomposition, which is why the fossil record is biased toward hard-bodied, rapidly buried organisms and is necessarily incomplete — soft tissues, rainforests, and low-population species fossilize poorly. Despite this bias, when fossils of a lineage are ordered by stratigraphic age, they show a directional sequence of trait change — for example, the stepwise reduction of hindlimbs and elongation of the body in the whale lineage over roughly 10 million years (Pakicetus → Ambulocetus → Rodhocetus → Basilosaurus).

Step 3 — Comparative anatomy reveals the "same bones, different jobs" pattern. Developmental gene networks (e.g., Hox genes patterning the limb axis) are conserved across tetrapods, so evolution modifies existing structures via changes in growth rate, bone fusion, and digit number rather than building new structures from scratch. This is why the pentadactyl limb persists — sometimes reduced (horse: one functional digit), sometimes re-elaborated (whale flipper: digit bones re-lengthened) — across mammals with vastly different lifestyles.

Step 4 — Vestigial structures are the fossil record's anatomical echo in living bodies. Genes for a structure can persist in a genome even after the structure loses its function, especially if the gene network is pleiotropically linked to other necessary functions. This explains functionless-but-present structures: the human coccyx (vestigial tail), the human appendix (a reduced cecum used for cellulose digestion in herbivorous ancestors), and pelvic and femur bones embedded in whale muscle with no connection to a leg.

Step 5 — Biogeography reflects dispersal and isolation, not a static "creation map." Speciation occurs when populations are geographically isolated (see Allopatric Speciation) and then diverge. This predicts that closely related species cluster geographically even when their habitats are ecologically similar to unrelated species elsewhere — Australian marsupials filling ecological niches analogous to (but not related to) placental mammals elsewhere is a textbook case of this pattern, sometimes miscalled "convergent evolution across a continent gap."

Step 6 — Molecular sequence comparison reconstructs the tree independently of morphology. DNA and protein sequences accumulate substitutions roughly clock-like over time (the "molecular clock," calibrated against fossil-dated divergences). Comparing homologous gene or protein sequences across species (e.g., cytochrome c, hemoglobin) lets researchers build a phylogenetic tree from sequence divergence alone — a tree that, when compared to the anatomically-derived tree, matches to a striking degree of statistical improbability if the two were generated by unrelated processes.

Fossil transitions: the tetrapod-to-whale and fish-to-tetrapod sequences. One of the most quantitatively precise fossil demonstrations is the whale lineage. Fossils dated by radiometric methods show, in strict chronological order: Pakicetus (~50 Ma, terrestrial, four functional legs, ear bones already showing early aquatic adaptation), Ambulocetus (~49 Ma, amphibious, large hind limbs used for swimming), Rodhocetus (~47 Ma, reduced hind limbs, likely could not support weight on land), Basilosaurus and Dorudon (~40–34 Ma, fully aquatic, tiny disconnected hind limb bones, elongated body). This ordered sequence is precisely what common descent with gradual modification predicts and precisely what "separately created, immutable kinds" does not predict. Similarly, Neil Shubin and colleagues in 2004 deliberately searched Devonian-age rock (~375 million years old) in the Canadian Arctic specifically because evolutionary theory predicted an intermediate between lobe-finned fish and early tetrapods should exist in rock of that age. They found Tiktaalik roseae: a fish with fins containing wrist-like bones, a mobile neck, and a flattened head — an animal with fish features (scales, gills) and emerging tetrapod features (weight-bearing limb bones, mobile neck) in exactly the rock layer theory predicted. This is a genuine predictive test, not a post-hoc rationalization: the theory specified where to dig and what should be found there, and the prediction succeeded.

Homology and Darwin's own use of comparative anatomy. Darwin devoted substantial space in On the Origin of Species to the argument that "unity of type" — the recurring pentadactyl limb pattern across mammals with radically different limb functions — makes sense under common descent but requires no explanation whatsoever under a model of independent, purpose-built design (why would a designer reuse an inefficient five-digit blueprint for a swimming flipper?).

Biogeography: Darwin's Galápagos observations and island endemism. Darwin noted during the Beagle voyage (1831–1836) that the Galápagos Islands, though ecologically similar to parts of the South American mainland, held finches, tortoises, and iguanas found nowhere else — but clearly related to mainland South American species rather than to similar species on similarly isolated islands elsewhere (e.g., in the Indian Ocean). Oceanic islands overwhelmingly lack native amphibians and terrestrial mammals, because these groups cannot survive prolonged salt-water crossing, while birds, insects, and some reptiles that can raft or fly do colonize such islands. This pattern of "disharmonic faunas" — islands missing whole classes of organisms that thrive on nearby similar mainland habitat — is a direct prediction of colonization-by-dispersal-then-divergence, and a direct problem for a model where organisms are placed wherever habitat suits them.

Molecular convergence: independent reconstruction of the same tree. Perhaps the most powerful modern evidence is that molecular phylogenies, built from DNA or protein sequence data using no anatomical information whatsoever, closely match trees built from morphology and fossils. A landmark example: comparisons of cytochrome c protein sequences across dozens of species (pioneered in the 1960s–70s by researchers including Margaret Dayhoff) showed that the number of amino acid differences between species' cytochrome c proteins scales with the time since their last common ancestor as independently estimated from fossils — humans and chimpanzees differ by 0 amino acid substitutions in this protein, humans and rhesus monkeys by 1, humans and horses by about 12, humans and yeast by about 40+. The near-universality of the genetic code itself (the same 61 sense codons specifying the same 20 amino acids across nearly all known life, with only minor documented exceptions such as some mitochondrial codes) is best explained as an inherited feature from a single ancestral coding system, since a code is otherwise an arbitrary convention with no functional reason to be shared across independently originated lineages.

A crucial genomic case: shared "broken genes." Humans, along with other haplorhine primates (but not most other mammals), cannot synthesize their own vitamin C because the gene for L-gulonolactone oxidase (GULO) is present but disabled by shared mutations. Different mammal lineages that lost this ability independently show different disabling mutations, but humans and other primates that lost it share the same disabling mutations in the same gene location — the signature expected if the mutation occurred once in a common ancestor and was simply inherited by descendants, not independently and coincidentally in each species.

Example 1 — Reading a molecular divergence table to infer relatedness. Suppose a researcher sequences a 500-base-pair mitochondrial gene fragment in five species and finds the following pairwise percent sequence differences:

Using a molecular clock calibrated at roughly 1% sequence divergence per 500,000 years for this marker (a hypothetical but realistic rate), we can reason: A and B differ by only 2%, indicating a very recent common ancestor (~1 million years ago) — they are likely sister species. C is more distantly related to the A/B pair (9% divergence, ~4.5 Ma), consistent with C splitting off before A and B diverged from each other. D is more distant still from the (A,B,C) clade (~17–18%, ~8.5–9 Ma), and E is the most divergent from everything (~22–24%, ~11–12 Ma), making E the outgroup. This produces the nested tree topology (((A,B),C),D),E) — exactly the kind of pattern that, when built independently from morphological characters (skeletal features, for instance), should match if common descent is true. If the morphological tree instead grouped E with A and B, that mismatch would represent a genuine anomaly requiring explanation (incomplete lineage sorting or hybridization, not falsification of common descent itself, but a signal that this particular data set needs more scrutiny).

Example 2 — Testing "separate creation" versus "common descent" against the fossil record with an asymmetric prediction. Consider a hypothetical stratigraphic column where a paleontologist finds, from oldest to youngest layers: single-celled fossils (~3.5 Ga layer), simple marine invertebrates (~540 Ma, Cambrian layer), jawless fish (~480 Ma), lobe-finned fish and early tetrapods (~375 Ma, matching the Tiktaalik find), early mammals (~200 Ma), and modern large mammals only in the most recent (< 5 Ma) layers. A skeptic proposes an alternative: "all fossils reflect a single global catastrophic burial event, with no true chronological signal, and hydrodynamic sorting by body size/density explains the apparent order." This alternative makes a specific, testable prediction: fossil order should correlate with density and shape (hydrodynamic sorting), not with derived anatomical complexity or radiometric age, and it predicts NO orderly nested anatomical intermediates should appear at consistent depths across unrelated geographic locations worldwide. In reality, radiometric dating of the rock layers (using different isotope systems in different labs, blind to each other's results) consistently corroborates the relative stratigraphic order, and the same anatomical sequence (fish-like → tetrapod-like → mammal-like) appears at consistent depths on every continent studied, independent of the fossils' size or density. The hydrodynamic-sorting hypothesis is falsified by this cross-continental, cross-method consistency, while the evolutionary prediction — that older rock holds anatomically more ancestral forms, and this pattern repeats globally because it reflects a real temporal process, not a local sorting artifact — is corroborated.

Charles Darwin assembled the anatomical and biogeographic evidence for common descent, but he could not have anticipated molecular biology; the person who did more than perhaps anyone to bring molecular data into the evidentiary case for evolution was biochemist Margaret Oakley Dayhoff. Working in the 1960s and 70s at the National Biomedical Research Foundation, Dayhoff compiled the Atlas of Protein Sequence and Structure, one of the first systematic databases of amino acid sequences across species. By aligning protein sequences like cytochrome c and hemoglobin across dozens of organisms, she and her collaborators built some of the first quantitative molecular phylogenetic trees — trees constructed entirely from sequence data, with zero reference to bones, embryos, or fossils. When these molecular trees were compared to the trees biologists had already built from anatomy and the fossil record over the preceding century, they matched to a degree of statistical improbability, providing exactly the kind of independent corroboration that a genuinely well-supported scientific theory should generate.

Understanding convergent evidence for common descent is not merely historical trivia — it underpins active biomedical and conservation science. Comparative genomics relies on evolutionary relatedness to choose model organisms: mice and humans share enough conserved gene sequence and pathway homology (both being mammals with a common ancestor ~90 million years ago) that mouse models remain central to drug development and cancer research. Phylogenetic forensics in public health uses molecular-clock methods, directly descended from the Dayhoff-era logic, to trace the origin and spread of pathogens — HIV's zoonotic origin from simian immunodeficiency virus, and the early spread patterns of SARS-CoV-2, were both reconstructed by building sequence-based phylogenies and cross-checking them against known outbreak timelines. Agricultural breeding and GMO safety assessment depend on knowing which genes are homologous across crop species to predict how an inserted trait will behave. Conservation biology uses biogeographic and phylogenetic evidence to identify evolutionarily distinct and globally endangered (EDGE) species — island endemics like Galápagos tortoises or Hawaiian honeycreepers are conservation priorities precisely because their isolated evolutionary history (the same biogeographic pattern Darwin documented) means their genetic and morphological diversity cannot be replaced if lost. Finally, vestigial-structure logic informs surgery and medicine directly: understanding the human appendix and coccyx as evolutionary remnants, rather than as organs with an undiscovered vital function, shaped historical debates about their surgical removal and continues to inform anatomical education.

Ask the tutor: How do biologists calibrate the molecular clock, and how much can mutation rate variation across lineages throw off divergence-time estimates?