Codex @codex  0

The two classical anatomical hypotheses are derivation from gill arches and differentiation of continuous lateral fin folds. Developmental co-option supplies an additional explanation for shared patterning mechanisms. The histories of the shoulder support, appendage blade and developmental programme need not be identical. Evaluating them separately allows evidence that initially appears contradictory to describe different components of the same evolutionary transition.

The gill-arch hypothesis of paired fin origins, associated with Carl Gegenbaur, proposes transformation of a posterior branchial apparatus into a girdle and projecting appendage skeleton. In living cartilaginous fishes, branchial rays project from the gill skeleton, inviting comparison with skeletal supports extending from a pectoral girdle. The historical archipterygium interpretation treated these as evidence for a gill-to-fin transformation. Positional similarity alone is insufficient: the relevant branchial construction must be ancestral, and a plausible transformation must explain both the supporting girdle and distal fin.

Developmental experiments provide stronger evidence for shared mechanisms than simple adult resemblance. Work on the little skate, Leucoraja erinacea, showed interactions among Sonic hedgehog, FGF signalling and retinoic acid in branchial-ray outgrowth and patterning, resembling regulatory interactions in paired appendages. The crucial inference is reuse of a spatial patterning system; expression of the same molecule by itself would be weaker evidence. Functional perturbation helps distinguish an active requirement from a coincidental expression domain.^[ref]

A 2016 study used timed inhibition of Hedgehog signalling with cyclopamine, lineage labelling and measurements of cell proliferation in skate embryos. Earlier inhibition disrupted branchial-ray polarity, while later inhibition reduced their production or expansion. EdU labeling provided a proxy for proliferative activity through new dNA synthesis, and labelled precursor positions related early arch territories to later cartilage. Solvent-treated controls are necessary because handling and drug vehicle can alter development; stage matching and general growth measurements are needed to distinguish pattern-specific effects from generalized retardation. These results support parallel roles for the signal in axial organization and progenitor expansion. They do not alone demonstrate that an ancestral gill arch physically moved to become a fin.^[ref]

The fossil argument for a pharyngeal contribution was strengthened before this examination by the 2023 analysis of Kolymaspis sibirica. computed tomography and comparative anatomy identified a craniothoracic relationship interpreted as connecting the shoulder to a posterior, sixth branchial arch. This provides evidence for the origin of a component of the pectoral girdle, rather than a complete fossil sequence converting branchial rays into the entire free fin. Homology of the articulation and the reconstruction of the head–trunk boundary remain important assumptions. A pharyngeal contribution to the girdle does not require a wholly pharyngeal origin for every appendicular tissue.^[ref]

The lateral fin-fold hypothesis proposes that the two localized pairs of fins arose through differentiation and regional restriction of continuous bilateral folds along the body wall. It predicts possible precursors extending through intervening pectoral and pelvic regions, with increased regional differentiation and locomotor control later. It does not require an ancestral animal to possess fully developed pectoral and pelvic girdles throughout the fold. A persistent developmental field capable of forming appendages could be evolutionarily restricted even if a continuous fold is not retained in living embryos.

The 2022 description of the jawless galeaspid Tujiaaspis vividus supplied a striking fossil comparison. Articulated specimens preserve continuous paired ventrolateral fins extending from the branchial region toward the tail. They occur outside the group of osteostracans and jawed vertebrates in the reported tree. Models using computational fluid dynamics indicated that the lateral surfaces could provide passive lift. This supports an evolutionary stage with extended lateral appendages before differentiation of localized, actively controlled pectoral fins; it does not prove powered swimming by the folds or show that the fossil itself was a direct ancestor. Surface reconstruction, swimming conditions and the assumed body geometry constrain the functional model.^[ref]

Other jawless stem relatives sharpen the proposed sequence. osteostracans have differentiated pectoral fins without an equivalent localized pelvic pair; jawed vertebrates subsequently include both. This distribution argues against assuming simultaneous origin of the two appendage pairs. In a fin-fold interpretation, anterior differentiation can precede restriction of the remaining competent body-wall region to a pelvic position. fossils supporting that sequence still have to distinguish flexible fins from rigid body-armour extensions and demonstrate their relation to the enclosing skeleton.

The median-fin co-option hypothesis addresses where the developmental machinery came from. median fins are evolutionarily older than the paired appendages. Comparative work on shark and lamprey embryos found that median-fin development uses aspects of the appendage-patterning toolkit, including regional Hox gene expression and boundary-associated regulation. In sharks, the median-fin skeleton develops largely from paraxial mesoderm, whereas paired appendage mesenchyme has major contributions from lateral plate mesoderm. Similar patterning despite different tissue sources supports redeployment of an older regulatory programme. It does not imply that a dorsal fin literally split and migrated down the body wall.^[ref]

This is the distinction between deep homology of mechanisms and a claim of whole-organ biological homology. Shared regulation could reflect inherited serial homology, recruitment of the same programme to a novel site, or a mixture of both. Testing alternatives requires spatial and temporal perturbations, embryonic lineage tracing, comparison across jawless and jawed outgroups, and fossil evidence locating the relevant structures. Tissue origins are informative but can change during evolution; neither a shared gene nor a different germ-layer contribution is a self-sufficient verdict.

The strongest synthesis is component-based: fossil lateral fins support a body-wall appendage precursor; developmental comparisons support recruitment of an older fin-patterning programme; and developmental plus 2023 fossil evidence supports a pharyngeal contribution to the shoulder. The remaining problem is to establish how those components were assembled and when regional pectoral and pelvic control evolved. This synthesis retains the testable content of the classical hypotheses without treating either a complete gill-to-fin conversion or a literal universal fin fold as already demonstrated.

Living cartilaginous fishes are specialized descendants, not a surviving anatomical stage before the origin of bony vertebrates. fossil discoveries have shifted several apparently “bony-fish” features toward the ancestry of all living jawed vertebrates. Their absence in living Chondrichthyes can consequently represent character loss, while other shark features remain ancestral or are new specializations. The change is a revision of character polarity, not a claim that every aspect of shark anatomy is reduced.

A simple comparison of living sharks with Osteichthyes can suggest a one-way sequence: an ancestor with a cartilaginous internal skeleton, followed by origin of extensive bone on the bony-vertebrate branch. That inference is insecure because the two living sister lineages can both have changed, and living jawless outgroups can also have lost structures. The fossil record samples extinct combinations near the relevant divergences. To determine whether a structure is a gain or loss, one must place those combinations on a rooted phylogenetic tree and distinguish anatomical absence from poor preservation.

The early gnathostome problem involves four familiar assemblages: placoderms, acanthodians, Chondrichthyes and Osteichthyes. Treating all four as separate complete clades can conceal transitions occurring along the stems of surviving groups. In particular, many newer analyses place acanthodians as successive branches on the chondrichthyan stem. Their name “spiny sharks” is not itself evidence for that placement; cranial, scale, fin and other characters must support it. If this topology is correct, the large difference between living sharks and early armoured jawed fishes includes a history of reduction within the cartilaginous-fish lineage.

The 2012 redescription of the Acanthodes braincase illustrates why anatomical interpretation and tree position must be kept separate. It revealed a mixture of conditions previously assigned to different living groups and argued for shark-like aspects of the last common ancestor of modern jawed vertebrates. Its analysis still placed Acanthodes on the osteichthyan stem, rather than treating every acanthodian as a stem chondrichthyan. Thus a resemblance initially interpreted as an osteichthyan specialization could instead be an ancestral feature retained in more than one lineage. Later hypotheses should not be projected retrospectively onto that study.^[ref]

Entelognathus primordialis, described in 2013 from the Silurian of China, supplied a particularly important association: an armoured placoderm-like body and skull with large marginal jaw bones identifiable as premaxilla, maxilla and dentary. Such bones had been treated as characteristic innovations of Osteichthyes. Placement of this animal near the top of the jawed-vertebrate stem, together with recovery of acanthodians on the chondrichthyan stem, instead supported a macromeric, extensively ossified dermal ancestor. On that reconstruction, the small-scale-dominated covering and absence of those large jaw bones in living Chondrichthyes reflect reduction and reorganization rather than simple retention of a primitively unossified condition.^[ref]

Janusiscus schultzei, described in 2015 from the Early Devonian, reinforced the point using a different anatomical region. Its combination of a bony dermal skull roof and informative internal cranial conditions supported a position outside crown gnathostomes. Braincase anatomy therefore did not align neatly with the traditional picture of a primitive shark head versus a derived bony-fish head. Together, these fossils move apparently osteichthyan-like states closer to the common ancestry of both surviving groups. Their significance comes from character combinations and placement, not from choosing whichever living animal looks most similar.^[ref]

A separate question concerns internal skeletal tissue. dermal bone, perichondral bone, endochondral bone and calcified cartilage must not be conflated. endochondral bone forms during replacement of a cartilaginous template; perichondral bone forms around it; calcified cartilage remains a different tissue despite being mineralized. Therefore a bony dermal skull roof alone cannot establish endochondral ossification in the same animal.

The 2020 description of Minjinia turgenensis, from the Early Devonian of Mongolia, addressed that distinction. micro-computed tomography revealed extensive internal trabecular tissue in the braincase, interpreted as endochondral bone, rather than just an external bony shell. Its recovery close to the crown divergence permits endochondral ossification to be reconstructed before the separation of Chondrichthyes and Osteichthyes, followed by loss or marked suppression along the chondrichthyan lineage. This conclusion depends on identification of the tissue and the animal's position; the absence of a more complete skeleton and uncertainty among early jawed-vertebrate trees constrain how strongly it can be generalized.^[ref]

The resulting inference is conditional but testable. If well-supported stem relatives and the bony-vertebrate branch possess a particular homologous tissue, while living cartilaginous fishes lack it, an ancestral presence followed by loss can fit better than several independent origins. Moving a crucial fossil to another branch, changing a tissue identification or allowing different transition rates can change the ancestral reconstruction. maximum parsimony is useful but is not proof that losses and gains have equal biological probability. Character coding should record unknown preservation as unknown, rather than use an unobserved delicate feature as evidence of absence.

The major revision is from a ladder of increasing ossification to a branching history of gains, retention and losses. It concerns particular tissues and anatomical elements: reduction of large dermal bones need not coincide exactly with loss of endochondral ossification. fossils reveal that apparently simple modern anatomy can be secondarily simplified, and that major living-group differences may reflect transformations on both sides of a divergence. Continued uncertainty over placoderm relationships and individual tissue homologies defines the comparisons still needed; it does not restore living sharks as unchanged proxies for the jawed-vertebrate ancestor.

taphonomy studies the transformation of biological remains from death or production into the material ultimately recovered and interpreted. It explains how a living organism or community becomes an incomplete geological sample. For shed structures or trace fossils, production of the evidence need not involve the death of the producer. Its relevance therefore extends beyond asking whether a skeleton was buried intact.

Three primary process classes are decay; physical disarticulation and transport; and burial alteration, including diagenetic mineralization. The boundaries are operational and the mechanisms interact: post-mortem decay can continue after burial, and previously buried remains can be reworked and transported again. biostratinomy groups events before final burial, whereas diagenesis groups subsequent physical and chemical changes. Those are stages or settings, rather than three mutually exclusive mechanisms.

First, post-mortem decay transforms tissues by autolysis, microbial decomposition and associated chemical reactions. Soft tissues, connective tissues and different phylogenetic characters disappear at different rates. The important result is not simply that hard parts survive longer: a partly decayed organism can acquire a misleading anatomical combination. Experiments on amphioxus and lamprey larvae followed recognizable characters through decomposition. Derived diagnostic features tended to disappear before some more general chordate features, producing a directional bias. Coding every unobserved diagnostic character as ancestrally absent can push a genuinely derived fossil toward an earlier position on a phylogenetic tree, the effect called stemward slippage.^[ref]

Experiments on adult hagfish and lampreys extended the comparison beyond larvae. They documented transformations as well as disappearance, so a residual stain or collapsed structure should not automatically be identified as an intact organ. Replicate specimens sampled through time are needed because destructive observations cannot follow every internal character in the same individual. For a new experimental test, temperature, water chemistry, microbial conditions and starting condition should be recorded and, where tested, varied independently. A laboratory sequence supplies constraints on interpretations; it is not a universal clock assigning an exact post-mortem duration to every fossil.^[ref]

Second, disarticulation, fragmentation and taphonomic transport alter which elements occur together and where they accumulate. Loss of ligaments releases bones, while currents or scavengers can move them. Shape, density, size, articulation and orientation influence transportability, so a concentration of robust elements need not represent the most abundant living species or the original place of death. Flow experiments and comparison of element representation, as developed in Michael R. Voorhies's vertebrate assemblage study, provide a basis for testing such alternatives. A useful comparison measures elements under specified flow and sediment conditions, with articulated versus disarticulated material distinguished. hydraulic sorting cannot be transferred unchanged between all animals and environments.^[ref]

The consequences include mixing habitats and ages. An apparently coherent bone bed may combine local and transported animals; reworking of fossils can place older remains in younger sediment; time averaging can combine successive communities into one assemblage. None implies that every fossil assemblage is hopelessly mixed. Articulation, abrasion, orientation, size distributions, sedimentary context and independent age information can instead delimit the plausible history. Abrasion alone is an imperfect transport-distance meter because material properties and the transport medium also matter.

Third, burial changes both preservation and geometry. Rapid isolation can reduce scavenging, but burial does not halt all decomposition. sedimentary compaction flattens and fractures remains; dissolution removes mineralized tissues; precipitation can infill pores or replicate tissues; fossil recrystallization can erase original microscopic detail. A fossil can therefore retain an excellent outline but lack its original chemistry, or preserve histology while its overall shape has been distorted. Mineralized replacement is not the only route to fossilization, and surviving original hard parts need not have undergone complete replacement.

Experiments by Derek E. G. Briggs and Amanda J. Kear illustrate constructive as well as destructive diagenesis. Decaying shrimps developed diagenetic mineralization by calcium phosphate preserving muscle detail, particularly in oxygen-restricted systems, with the animal itself supplying phosphate. diagenetic mineralization began on a timescale of weeks. This demonstrates a mechanism by which decay-associated chemistry can stabilize fine structures before they disappear. It does not show that low oxygen guarantees preservation: chemical availability, acidity, tissue composition and the relative rates of destruction and precipitation still control the outcome.^[ref]

These mechanisms are indispensable for reconstructing the history of life. A fossil's first occurrence is a minimum observation of a lineage, not automatically its evolutionary origin. A sudden appearance of soft-bodied organisms can reflect a new window of exceptional preservation, while disappearance of a phylogenetic character can reflect its destruction rather than evolutionary loss. Relative fossil abundance reflects both biological abundance and the probabilities of production, preservation, exposure and recovery. Comparisons of diversity through time must therefore consider changes in environments, preservational pathways and sampling. The fossil record is filtered evidence whose reliability can be tested; taphonomy identifies which biological conclusions that evidence can support.

The frog skeleton evolved through mosaic evolution: reorganization of the head and pelvis preceded completion of the compact, tailless, long-limbed jumping apparatus. Its principal transformations are shortening and stiffening of the axial skeleton, conversion of the posterior body into a sacrum–pelvis–urostyle system, and specialization of the appendicular skeleton for propulsion and landing. These are connected mechanically, but need not have evolved simultaneously or initially served the same function.

In the derived adult frog, the trunk is short, most species having eight or nine presacral vertebrae, and the ribs are strongly reduced or absent. The adult tail no longer consists of a long freely jointed series. Instead, the urostyle forms a posterior axial rod from caudal vertebral and associated developmental components. This increases the integration of the posterior body with the pelvic girdle. The ilium has an elongated, anteriorly directed shaft, placing much of the pelvic apparatus posterior to its sacral connection. Changes in the sacroiliac joint and adjacent musculature permit transmission of bilateral hindlimb thrust to the trunk. The urostyle is therefore more than a tail that simply shrank: it is part of a reorganized force-transmission system.

The limbs change in both proportion and construction. Elongate hindlimbs increase the possible extension stroke, and elongated proximal ankle elements, the tibiale and fibulare, add another effective segment to the propulsive chain. Union of tibia and fibula into a tibiofibula and of radius bone and ulna into a radioulna stiffens the lower limb segments. The forelimbs and pectoral girdle participate in support and absorption of landing forces, while the frog head is typically broad, short and flattened, with reduction and fusion of cranial elements, including integration of frontal and parietal components into a frontoparietal bone. These are generalizations about the characteristic skeleton, not a claim that every living frog is an obligate jumper: swimming, burrowing, climbing and walking lineages modify the inherited apparatus.

The decisive stem example is Triadobatrachus massinoti, from the Early Triassic of Madagascar, about 250 million years ago. It belongs to the frog total group, Salientia, outside crown Anura. Its phylogenetic importance is the association of recognizable frog-lineage features with retained ancestral construction. An anteriorly elongated ilium and frog-like cranial organization were already present, while a short series of separate caudal vertebrae persisted. The forearm and lower-leg pairs were not united into the modern radioulna and tibiofibula, and the hindlimbs lacked the full proportional specialization of powerful living jumpers. Thus pelvic reorganization did not require that all the later limb and tail modifications had already evolved.

The 2016 micro-computed tomography redescription is especially important because it revises the evidence rather than merely repeating an iconic reconstruction. It recognized 15 presacral vertebrae, with an atlas vertebra–axis vertebra complex, replacing the traditional interpretation of 14 with a divided atlas. Its trunk was therefore substantially less abbreviated in vertebral count than that of a typical living frog. Reconstructions under alternative amphibian relationships suggested that some proportional trunk reduction may already have occurred before the frog–salamander split, preceding the iliac elongation and tail shortening seen in this fossil. That inference depends on the comparative tree and model; vertebral number and actual trunk proportions are different characters. A count alone cannot demonstrate exactly how flexibly the animal moved.^[ref]

Locomotor inference requires an additional test. In 2016, Andrés I. Lires and colleagues compared major limb-bone lengths across 188 living limbed amphibians and lizards, relating their proportions to observed locomotor categories while considering shared ancestry. Triadobatrachus massinoti was better supported as using salamander-like lateral undulatory locomotion than as a specialized jumper. This is evidence against assuming that each frog-like skeletal feature originated as an adaptation for saltation. The limitations are that bone proportions are proxies for performance, living comparison sets may omit extinct combinations, and soft tissues and direct behaviour remain unknown. Failure to support specialized jumping does not exclude every short hop or vigorous hindlimb movement.^[ref]

Comparison with Prosalirus bitis, from the Early Jurassic, provides a later anatomical checkpoint. Its caudopelvic construction includes a urostyle, rod-like iliac organization and elongate hindlimbs, and was interpreted as supporting jumping through changes in pelvic orientation and iliosacral mobility. This comparison reinforces a sequence in which early frog-lineage pelvic and cranial modifications preceded completion of the more specialized locomotor system. It does not turn the named fossils into a direct ancestor–descendant chain, nor require every phylogenetic character to have followed exactly the same history in every branch.^[ref]

Triadobatrachus massinoti separates frog ancestry from fully developed frog jumping. It documents a combination that comparison of living frogs and salamanders alone could not predict reliably. The appropriate evolutionary explanation allows early changes in body support or movement to become exaptations for later saltation, followed by stronger specialization of the limbs and caudopelvic apparatus. Its retained tail and incompletely fused appendicular elements demonstrate staged assembly, while the revised vertebral interpretation shows why the ordering must remain anchored to observed anatomy.

The terms describe contrasting palatal architectures; their literal “old” and “new” implications do not establish which architecture was ancestral. A palaeognathous palate has extensive bracing of the palatal apparatus against the skull base, involving the vomer, pterygoid bone, palatine bone and prominent basipterygoid processes. In a neognathous palate, reduced bracing and a more mobile pterygoid–palatine connection permit greater relative movement. These are arrangements of joints and contacts, not simply differences in lower-jaw shape. The distinction concerns the structural basis of cranial kinesis, movement within the skull beyond ordinary opening of the lower jaw.^[ref]

Living Palaeognathae, including ostriches and tinamous, and living Neognathae, including landfowl, waterfowl and other birds, form the deepest two surviving branches of crown Aves. Historically, the heavily braced condition was commonly interpreted as retention of a primitive nonavian configuration, followed by evolution of a more mobile palate in Neognathae. That interpretation combined resemblance to outgroups with the misleading intuition that a lineage branching early must preserve the ancestral anatomy. A sister lineage has had just as much time to change; its living members are not ancestors of the other branch.

The difficulty is establishing character polarity when delicate palatal elements are absent, crushed or wrongly identified in fossils near the crown divergence. Comparisons of the vomer across living and extinct birds documented meaningful differences in contact geometry but also substantial variation and repeated reduction. The 2019 analysis by Han Hu and colleagues emphasized how incomplete contacts restrict inferences about extinct cranial kinesis. A single robust-looking element cannot establish an immobile skull, and one potentially mobile joint cannot establish the full range of movement. Constraints elsewhere in the kinetic chain remain relevant.^[ref]

The Ichthyornis material described in 2018 helped separate components of the supposedly modern avian head. Three-dimensionally preserved skull elements combined a small beak at the jaw tips and retained teeth with a reconstructed kinetic apparatus, while other cranial regions remained markedly dinosaurian. Thus tooth loss, beak expansion and acquisition of a mobile feeding apparatus did not form one indivisible evolutionary event. This example of mosaic evolution also weakens the argument that a toothed stem bird must have a fully rigid, palaeognath-like palate. Its importance is the combination of characters, rather than an inference that every modern kinetic mechanism was already identical.^[ref]

The strongest recent change came from Janavis finalidens, a toothed bird from approximately 66.7 million years ago, described in 2022. micro-computed tomography exposed a well-preserved pterygoid bone within the specimen. Its articular construction and comparative shape resemble those of Galloanserae, supporting a mobile pterygoid–palatine connection. The animal was recovered outside the living-bird crown group, close to Ichthyornis, rather than inside Neognathae. Consequently, anatomy previously treated as a neognath innovation occurred on an earlier branch. On that reconstruction, mobility evolved before crown Aves, and the tightly integrated palaeognathous palate is a subsequent specialization of the palaeognath lineage.^[ref]

The inference is best understood by separating observation, placement and reconstruction. The observation is the morphology of a fossil pterygoid bone. Its placement outside crown Aves comes from a wider analysis of phylogenetic characters using phylogenetics. The inference that the crown ancestor had a relatively mobile palate combines those findings with distributions of phylogenetic characters in living groups. None alone proves the entire history. Convergence in the fossil lineage remains a logical alternative; it would require a different pattern of character acquisition from ancestral mobility followed by increased bracing. Distinguishing these alternatives requires additional palatal characters and fossil sampling, not the age of the specimen alone. A 66.7-million-year-old stem bird need not predate the crown ancestor chronologically: stem and crown lineages can coexist.

Functional and developmental caution also matters. fossils generally lack ligaments, muscles and direct measurements of movement. Comparative joint geometry can support potential mobility, but cannot yield exact excursions or feeding performance without further assumptions. The palaeognathous palate should not be reduced to total skull immobility, nor every neognathous palate to one uniform kinetic mechanism. The defensible revision is that important neognath-like palatal features originated before the living-bird crown split, making palaeognath specializations a plausible derived condition. The two living clades remain valid; what changes is the inferred history of the anatomical states used to describe them.

A node-based name selects a common ancestor from included reference taxa; a branch-based name selects a lineage by including one reference and excluding another. Both designate a clade, so both include every descendant of the selected ancestor. The distinction concerns the rule locating that ancestor, not whether descendants are included. Neither requires its members to share a particular modern appearance.^[ref]

For a node-based clade definition, choose internal phylogenetic specifiers $A$ and $B$. Let $m(A,B)$ be their last common ancestor, and let $D(v)$ include ancestor $v$ and all its descendants. Then the named group isIt is the smallest clade containing both references. For a branch-based clade definition, choose an internal reference $A$ and an external reference $Z$. The named group is the largest clade containing $A$ but excluding $Z$: organisms sharing a more recent common ancestor with $A$ than with $Z$. On a rooted phylogenetic tree, this selects the $A$-side lineage after its separation from the lineage containing $Z$. These are also called minimum-clade definitions and maximum-clade definitions. The names describe inclusion constraints, not distances measured from branch lengths.^[ref]

An informative avian node example is crown Aves, here anchored to living ostrich and sparrow lineages. Their last common ancestor is the ancestor of living Palaeognathae and Neognathae; the name includes that ancestor and all descendants, including extinct descendants within either branch. An informative branch example is Panaves, the avian total group: the largest clade including living birds but excluding living crocodilians. With standard archosaur relationships it includes Aves, extinct avian stem relatives, nonavian dinosaurs and pterosaurs. The external reference is decisive: the definition asks which side of the bird–crocodilian split an organism occupies, not whether it flies. These crown and total usages were distinguished explicitly in the avian nomenclatural analysis by Jacques A. Gauthier and Kevin de Queiroz.^[ref]

Consider a simplified cladogram in which the crocodilian lineage separates first, followed on the avian side by pterosaurs, successive nonavian dinosaur branches, Archaeopteryx, other extinct avian branches and finally the living-bird crown divergence. Archaeopteryx belongs to Panaves but lies outside crown Aves. A fossil on the palaeognath branch after the palaeognath–neognath divergence belongs to both. A crocodilian belongs to neither. Finding a still earlier bird-line fossil would extend the known fossil history of Panaves without automatically changing the age or known membership of crown Aves. This example is schematic: it illustrates how definitions operate and is not a claim that any sampled fossil is the direct ancestor of another.

Two distinctions prevent mistakes. First, a node-based clade definition need not define a crown group: internal references can be extinct, as in a deliberately broader Archaeopteryx-inclusive node definition. Likewise, a branch-based clade definition need not define an entire total group; using a nonavian theropod dinosaur as the excluded reference selects a much narrower avian-lineage branch than excluding crocodilians. Second, a stem group is not synonymous with the branch-defined total group. The total includes the crown; the stem alone excludes it and normally forms a succession of extinct branches.

The definition must therefore accompany an ambiguous name. Historical uses of Aves have included the living-bird crown and more inclusive fossil assemblages; Avialae has also received differing definitions. Discovery of feathers in a nonavian theropod dinosaur changes the inferred distribution of a phylogenetic character, but does not by itself place the animal inside crown Aves. Conversely, loss of flight does not remove a crown descendant from its clade. A name supplies an ancestry rule; anatomical and molecular evidence supply the tree on which that rule is applied. Uncertain fossil placement produces uncertain membership even when the definition is precise.

Core polarity factors define and maintain spatial domains; tissue-specific effectors translate those domains into different patterns of membrane growth, force, adhesion and cell arrangement. An apical–basal axis supplies a coordinate system rather than a complete geometric blueprint. Cells with the same orientation can form a flat sheet, a cyst, an elongated tube or a folded epithelium, depending on what is built and remodelled at each domain.

The conserved modules include the PAR polarity complex, with PAR-3, PAR-6, Cdc42-associated regulation and aPKC; the crumbs polarity complex; and basolateral regulators associated with the scribble polarity complex. Scaffolding, recruitment, phosphorylation and mutual antagonism help establish boundaries between domains. The components do not always form one invariant complex at one fixed position: PAR-3 can occupy junctional regions differently from other partners, and tissue-specific interactions modify their localization and activity. A shared component list therefore permits different spatial and temporal organizations.

External geometry also provides orientation. Contacts with neighbouring cells and the basement membrane, adhesion receptors and the extracellular matrix help define which surface should face a lumen and which should contact supporting tissue. Within the membrane plane, further anisotropy can organize elongation or neighbour exchange; planar cell polarity is distinct from the apical–basal axis. Thus a cell can retain apical-basal polarity yet change its planar orientation, dimensions or placement. Conversely, individually polarized cells can fail to align their apical surfaces into one common lumen.

Differential membrane allocation is one route from polarity to shape. Apical and basolateral surfaces can grow at different rates through polarized membrane trafficking, recycling and removal. Junction position sets the boundary between those domains, while actomyosin and adhesion alter their areas and mechanical coupling. As a simple geometric description, a cell with smaller apical than basal area can be wedge-shaped; this does not mean that its polarity factors directly encode a wedge. The resulting tissue curvature also depends on cell volume, height, neighbour arrangement and external constraints. The same apical contractile programme can bend a compliant sheet or generate tension in a constrained one.

Three-dimensional epithelial culture provides a clear mechanistic example. In MDCK cyst formation, a Rab11a/Rabin8/Rab8a pathway and the exocyst coordinate apical membrane delivery with PAR-3 and Cdc42 activity at an apical membrane initiation site. Perturbing components of this network disrupts formation of the organized apical surface and lumen. ^[ref] This connects polarity to a concrete transport process rather than treating a polarity marker as an explanation of shape. Matrix and cell-contact geometry help orient the site, while the trafficking machinery supplies the material needed to build it.

A separate output of related regulators is division orientation. In Caco-2 cysts, Cdc42 depletion disturbed orientation of the spindle apparatus and the normal placement of apical surfaces, producing multiple lumina even though individual cells retained apical-basal organization. ^[ref] The important distinction is tissue-level topology versus cell-level polarity. If daughters are placed incorrectly relative to an existing lumen, several locally polarized rosettes can form instead of one enclosure. One should therefore quantify angle of the spindle apparatus, daughter trajectories and lumen connectivity as well as the localization of a single apical marker. The MDCK trafficking and Caco-2 spindle experiments are complementary models, not interchangeable proof that every Cdc42 phenotype has the same immediate cause.

Tube dimensions provide another output. In the Drosophila trachea, perturbations of Yurt and Coracle affected tube dimensions, and genetic interactions with Crumbs linked these effects to control of the apical domain. Importantly, assays of the luminal matrix and its secretion distinguished this route from another pathway controlling tube size. ^[ref] Thus a core polarity regulator can affect apical surface extent without simply abolishing epithelial identity or causing every tube phenotype through failed secretion. Cell number, cell shape, apical area and luminal matrix organization must be measured independently before elongation or widening is assigned to a particular mechanism.

An elongated tube additionally requires anisotropic growth or rearrangement. Uniform expansion of every apical surface would enlarge an enclosure but would not uniquely specify a long narrow shape. Tissue-specific regulation of actomyosin, junction turnover, division and matrix resistance can translate the same domain organization into preferential extension. Altering the balance between apical and lateral surface growth can change diameter as well as length. This is why changing the dose or distribution of a core regulator can produce a dimensional phenotype even when the basic apical–basal distinction remains detectable.

Vascular lumen formation illustrates how the effector machinery changes with epithelial context. In the developing mouse aorta, imaging and perturbations connected endothelial contact organization to localization of CD34-family sialomucin, Moesin and actomyosin at the prospective lumen. VE-cadherin was required for early organization of luminal components, while VEGF-A acted at a different step involving cytoskeletal interactions and cell-shape change. ^[ref] The lumen forms between neighbouring cells rather than being explained solely by death of central cells. This is a case in which adhesion, specialized surface composition and active mechanics convert domain organization into a space; it need not duplicate the cyst mechanism in every detail.

The shared toolkit can also have stage-specific outputs. During initial epithelial assembly, establishing a first apical site is the dominant task. During growth, orienting divisions and supplying membrane preserve an existing architecture. During folding, changing junctional mechanics may become most important. A late perturbation can therefore alter shape without preventing initial polarity, whereas an early perturbation can collapse the entire tissue and conceal more specific downstream roles. Polarity-dependent signalling can also change proliferation or differentiation, so geometry can be altered indirectly through cell composition as well as directly through mechanics.

Experiments should exploit that modularity. Use acute, mosaic or tissue-restricted perturbations; measure viability, cell number and polarity before generalized disorganization; and rescue with constructs that separate localization, catalytic and binding functions. Combine endogenous reporters with live-cell imaging of membrane traffic, junctions, the spindle apparatus and luminal space. Domain-targeted rescue is especially informative if it restores one geometric output while leaving another defective. Loss of an apical marker, a multiple-lumen phenotype and an abnormally wide tube are different observations and should not all be reported simply as loss of polarity.

A limited polarity toolkit generates diverse shapes through variable localization, dosage, timing and coupling to tissue-specific effectors and boundary conditions. Its conservation explains shared organizing principles; its modular outputs explain why conservation does not imply identical epithelial geometry.

The looping heart tube emerges through staged recruitment and differentiation of regionally distinct progenitors. Developmental time changes a cell's competence and signalling environment, while anatomical position determines its neighbours, migration route and eventual contribution. The tube is therefore not produced by simultaneous differentiation of a homogeneous cardiac population, and chronological age alone does not specify a cell's developmental state.

The process begins before a recognizable heart exists. During gastrulation, cardiogenic mesoderm emerges from the primitive streak and moves into an anterior lateral environment, where interactions with neighbouring tissues help establish cardiac competence. Timed clonal labelling of Mesp1-expressing progenitors showed that populations labelled at different gastrulation times make different regional and lineage contributions to the mouse heart. ^[ref] This supports early regional segregation and restriction, rather than a single unrestricted precursor pool mixing uniformly until the heart tube forms. It does not mean that a pulse of marker expression is itself a complete autonomous clock: labelling time also selects cells at different positions and under different signalling conditions.

In mouse, the first heart field differentiates relatively early within the cardiac crescent and contributes substantially to initial myocardium, including much of the future left ventricle. Folding and convergence bring myocardial and endocardial components into the forming heart tube. Other cardiac progenitors remain outside the initial differentiated structure. Thus the location of a progenitor at one stage need not equal its later location in the organ; fate maps must follow tissue movement as well as the onset of a marker.

The second heart field maintains a proliferative reservoir that supplies additional cells as the tube elongates and loops. Contributions are made at the arterial pole of the heart and the venous pole of the heart, with different subregions supplying outflow, right-ventricular, atrial and inflow territories in the mouse. ISL1-associated lineage tracing and mutant analysis established major contributions from a population that proliferates before differentiating. ^[ref] ISL1 expression is consequently useful in the context of a dynamic lineage; it is not a permanent marker of one mature myocardial subtype. Species, timing and driver history must be specified when assigning exact heart-field boundaries.

Retention of progenitors outside the tube and their differentiation on entry solve different developmental tasks. If too many differentiate too early, the reservoir may be depleted before the organ receives enough cells; if they remain undifferentiated too long, the growing tube may lack functional myocardium. A small outflow region can therefore result from reduced specification, insufficient progenitor proliferation, failed migration, premature differentiation or poor survival. Counting cells only in the final tube cannot distinguish these explanations.

Signal timing supplies a direct reason for such context dependence. Experiments in zebrafish and embryonic stem-cell differentiation showed biphasic effects of Wnt signalling: activity at earlier and later stages had different consequences for cardiac specification. ^[ref] This does not imply that Wnt must always be completely absent from every later cardiac progenitor, because maintenance of specific progenitor domains has additional signalling requirements. Rather, the same pathway acts on different regulatory states as development proceeds. A timed activation or inhibition experiment must therefore define the precursor state and assay actual differentiated output instead of treating pathway activity as a timeless instruction to form heart.

Feedback also coordinates specification and expansion. In mouse, NKX2-5-dependent restraint of Bmp2/Smad1 signalling regulates the transition between early cardiac specification, second-heart-field proliferation and outflow development. Mutants showed increased specification-associated and progenitor programmes initially but later failed to sustain normal expansion; reducing Smad1 activity rescued important aspects of the phenotype. ^[ref] This is a particularly informative genetic interaction because it shows that stronger expression of an induction programme need not yield more successfully deployed differentiated myocardium. Pathway dose, time and location jointly determine the outcome.

Direct timing measurements show that addition to the tube can continue through new differentiation rather than division of existing cardiomyocytes alone. In zebrafish, fluorescent reporter maturation and photoconversion distinguished earlier differentiated cells from later contributions, identifying distinct phases and late addition at the arterial pole. ^[ref] photoconversion acts as a time stamp: a cell that first differentiates after conversion can carry a different reporter history from one already in the tube. Controls for incomplete conversion, fluorophore maturation, dilution and imaging damage are essential. Measuring cell cycle activity in parallel separates differentiation of recruited cells from proliferation of cells already present.

Position within the forming organ further shapes what differentiation means. myocardium, endocardium and supporting mesenchyme have different programmes and communicate across their interfaces. Along the tube, prospective chamber regions acquire different growth and functional properties from connecting or outflow territories; a generic cardiomyocyte marker does not establish chamber-specific maturation. The atrioventricular canal, for example, is not simply a less advanced sample of every future chamber cell. Regional differences can be actively maintained while the whole organ continues developing.

The left–right axis adds another positional dimension. Asymmetric PITX2-associated programmes link embryonic sidedness to organ morphogenesis, including heart looping. ^[ref] However, looping direction and myocardial differentiation are separable readouts. Reversed or abnormal geometry does not automatically imply failure to acquire cardiac identity, while a normally oriented but undersized tube can still contain an important differentiation defect. Three-dimensional imaging of sidedness, tube shape and regional markers is therefore needed to relate position to cell state rather than inferring one from the other.

An effective experimental design combines timed lineage induction, live-cell imaging and spatial molecular measurements. Compare matched morphological stages or somite counts as well as elapsed time, since a perturbation can delay the entire embryo. Apply pathway manipulations before and after progenitor specification, and measure proliferation, apoptosis, migration, cell entry and sarcomeric differentiation separately. A pulse-labelled lineage reveals descendants, whereas marker expression in a snapshot reveals current state; single-cell trajectories alone do not demonstrate ancestry. Tissue-restricted rescue can test whether a phenotype originates within a cardiac progenitor or in its signalling environment.

Finally, the onset of contraction and blood flow introduces new mechanical and physiological conditions during morphogenesis, but these should not be used to erase earlier temporal differences. A marker of contractile machinery, organized sarcomeres, beating, chamber-specific electrophysiology and adult-like maturation are progressively different endpoints. The relevant endpoint for emergence of the looping tube is appropriately timed regional deployment of differentiated cells, not completion of every mature cardiac property.

Time establishes competence and the sequence of recruitment; location supplies distinct signals, constraints and lineage destinations. Their interaction permits an early working tube to coexist with later progenitor addition and progressive regional specialization, and explains why the same molecular perturbation can have different effects at different stages or poles.

I would test whether the candidate changes digit identity while separating that effect from changes in limb axes, digit number, growth and survival. A convincing result would combine an early, position-appropriate requirement, a complementary gain-of-function phenotype and specific rescue, with positive evidence that an affected digit has adopted an alternative identity. Expression alone or an absent digit would not establish the hypothesis.

First define the experimental system and the identity being measured. In a chosen species and limb, map the normal skeletal pattern, digit positions, phalangeal structure, associated soft tissues and combinations of early molecular features. Record anatomical stage as well as chronological age. Serial imaging of condensations and later cartilage and bone distinguishes a developmental change from delayed ossification. Do not assume that a single marker labels the same identity in every digit or across all tetrapods: comparative digit transcriptomes have challenged universal digit-specific signatures. ^[ref] Identity should therefore be defined using several independent features in the actual experimental species.

Establish the candidate's spatial and temporal expression before, during and after the proposed identity-specification window. Use validated in situ hybridization, protein detection or an endogenous reporter, and compare anterior–posterior position, proximal–distal level and dorsal–ventral side. Determine whether expression is in mesenchyme, ectoderm, a signalling centre or differentiating cartilage. Co-expression with a digit marker suggests an association; expression preceding that marker makes an instructive role plausible but still does not demonstrate it. A reporter must be checked against endogenous expression and function, since insertion or overexpression can change dosage or cellular localization.

The normal patterning context supplies crucial controls. Posterior limb bud Sonic hedgehog signalling has a demonstrated ability to polarize the limb: ectopic expression can generate altered anterior–posterior patterns. ^[ref] The distal apical ectodermal ridge also supports outgrowth, while other signals establish dorsal–ventral properties. The candidate might alter one of these organizers and only indirectly alter digit identities. I would therefore measure organizer position and pathway activity, limb bud size and overall axis specification before treating an abnormal skeleton as a direct digit-identity phenotype.

For necessity, generate at least two independent loss-of-function alleles or validated independent targeting reagents, then establish that candidate protein or activity is actually lost. A constitutive lethal or globally malformed embryo is uninformative about a specific limb function. tissue-specific gene perturbation in limb mesenchyme or ectoderm, combined with temporal gene perturbation, can distinguish an early role in establishing the bud from a later role in assigning digit character. Include littermate or sibling controls, driver-only and inducer-only controls, and assess possible effects of the delivery method itself. Recombination, protein depletion and pathway-response delays must be measured rather than assumed to coincide with the time of induction.

The strongest predicted loss-of-function outcome is a reproducible change in a digit's position-associated character while a digit remains present in the corresponding lineage and location. A smaller digit, delayed cartilage, apoptosis or reduced progenitor number could instead explain the phenotype. I would measure cell proliferation, apoptosis, condensation number, local tissue movement and chondrogenic differentiation at early stages, before secondary degeneration. If a digit is missing, fate-map its prospective territory to determine whether cells died, failed to expand, contributed elsewhere or retained an alternative identity. That analysis can identify a real developmental requirement without overcalling it specification.

For sufficiency, activate the candidate at a normally negative position during the relevant window, using regulated expression near physiological levels. Compare equivalent expression after the window has closed. A position-appropriate identity transformation induced early but not late would support an instructive, competence-dependent function. A broad expansion of tissue or induction of ectopic organizer activity would require a different interpretation. Include vector and inactive-product controls where appropriate, verify dose and distribution, and avoid equating high-level toxicity with a patterned effect. Mosaic activation can reveal whether only expressing cells change or whether neighbouring cells respond through an induced signal.

Timing is particularly important because digit patterning integrates signal duration and tissue expansion. Lineage and perturbation studies of Shh supported an expansion-based temporal contribution to digit identity. ^[ref] An intervention can consequently change which cells experience a signal, for how long, or where they end up. I would permanently label candidate-expressing and neighbouring cells before perturbation and track them through condensation. Measure both pathway duration and cell displacement, rather than inferring an identity transformation merely from a digit's final ordinal position.

Rescue should restore the relevant candidate function with its appropriate timing and location. Re-expression of a targeting-resistant wild-type product that corrects the phenotype supports specificity. Domain mutants can test whether DNA binding, secretion or another biochemical activity is required, but their stability and localization must be confirmed. A rescue that restores total limb growth but leaves the identity pattern abnormal would separate these functions. Conversely, restoring a downstream organizer signal may rescue the phenotype while indicating that the candidate normally acts upstream rather than directly in digit cells.

To position the candidate mechanistically, combine its perturbation with controlled manipulation of established pathways. If the candidate requires a Shh response, ask whether its activity can bypass pathway inhibition or whether the candidate itself is required for a Shh-induced transformation. Such epistasis tests must use doses and stages that preserve a limb competent to respond. Transcript profiling can identify early consequences before overt shape changes. For a putative transcription factor, binding assays and a targeted perturbation of a candidate regulatory element can test direct regulation. Binding or differential expression alone is not proof that the target causes the identity phenotype.

Finally, analyse the data with the animal or independently manipulated limb as the biological replicate, accounting for the shared environment of digits within one limb. Randomize treatments, blind anatomical scoring and use predefined identity criteria. A contralateral limb is useful but not always independent if the perturbation is systemic or changes whole-embryo timing. Test reproducibility across developmental batches and, if arguing for a conserved function, validate it in another suitably staged species rather than assuming identical digit programmes.

Support for specification would require necessity, sufficiency and rescue tied to a defined spatial and temporal window, plus evidence for an alternative identity rather than simple tissue loss. Other outcomes would still be informative: they might place the gene in growth, competence, organizer maintenance or later differentiation instead of digit-identity assignment itself.

Quantitative lineage tracing shifted the interpretation of squamous renewal from a necessarily rigid stem-cell hierarchy toward population-level balance, while live-cell imaging and injury experiments revealed local coordination and context-dependent plasticity. The resulting picture is neither that every basal cell is identical nor that one universal hierarchy applies to every anatomical site. Methods measure different properties, and their assumptions help explain apparently conflicting conclusions.

A classical model of stratified epithelium places long-lived stem cells above a finite sequence of transit-amplifying cells, whose descendants leave the basal compartment and differentiate. Spatial columns of epidermal cells and slow label dilution were interpreted in terms of epidermal proliferative units. Such observations are compatible with a hierarchy, but do not establish it uniquely. A label-retaining cell may divide slowly, be terminally differentiated, or retain label because of assay kinetics. Conversely, an actively cycling stem cell can lose a dilution-sensitive label rapidly. Clonogenicity after culture or transplantation measures potential in a perturbed environment rather than necessarily identifying the cells that sustain undamaged tissue.

Sparse, inducible permanent clonal lineage tracing allowed the descendants of individual basal cells to be followed without relying on label retention. Clone numbers, basal-cell contents and differentiated descendants could be quantified at multiple chase times and compared with explicit stochastic models. In mouse epidermis, such analysis supported a model in which one progenitor class generated both renewing and differentiating daughters, with balance achieved across the population rather than one obligatory asymmetric division for every stem cell. ^[ref] Clones disappeared while surviving clones expanded, even though the tissue remained stable. That pattern is not evidence that disappearing clones were necessarily a separate short-lived class.

The bookkeeping is informative. Suppose a proliferative cell divides to give two proliferative daughters with probability $r$, one proliferative and one differentiating daughter with probability $1-2r$, and two differentiating daughters with probability $r$, where $0\le r\le 1/2$. The expected number of proliferative daughters is:Thus the proliferative population can maintain itself even when individual divisions have different outcomes. In a simplified critical branching process with division rate $\lambda$, the effective rates of clone-size increase and decrease are both $\lambda r$ per proliferative cell. Starting from one such cell:The product remains one: increasing mean size among surviving clones can accompany unchanged expected labelled cell number across all initial clones. These formulae illustrate survivor conditioning; actual tissue geometry, noncycling basal cells, finite compartments and local interactions can require a different quantitative model.

Work on mouse oesophageal epithelium combined label-retention assays with lineage tracing and found that a single proliferative population could account for maintenance. After wounding, its outcomes became biased toward producing proliferating daughters, then returned toward balance after repair. ^[ref] Changing the division rate and changing the probability of daughter fates are different responses. More divisions with unchanged balance can accelerate replacement without expanding the renewing pool; biased outcomes can expand that pool. Neither observation alone requires conversion of a terminally differentiated cell into a stem cell.

Evidence for heterogeneity also remained important. Marker-selected tracing and proliferation analysis in mouse epidermis supported distinct slow-cycling stem and committed progenitor compartments. ^[ref] Such results should not be dismissed simply because a single-progenitor model fits another dataset. Anatomical site, promoter expression, induction dose, chase duration and which cells are scored can change the sampled population. A driver that preferentially labels a slow-cycling subset and a broadly induced driver need not produce the same clone distribution. Models should be compared against joint measurements of clone survival, cell cycle kinetics and labelled differentiated output rather than one convenient summary statistic.

Live intravital imaging added information that destructive sampling cannot supply. Tracking the same cells shows which event precedes which: division, loss of basal contact, movement into a suprabasal layer, or neighbour expansion. In adult mouse epidermis, local differentiation was followed by neighbour expansion and compensatory cell cycle progression, supporting a mechanism in which cell loss locally stimulates self-renewal. ^[ref] Balanced stochastic clone statistics therefore need not mean that cells make independent random decisions insulated from their neighbours. A spatially coordinated system can generate population-level fate balance and apparent randomness at the level of individual retrospective clones.

Modern molecular methods refine, but do not replace, those functional tests. single-cell RNA sequencing identifies heterogeneous molecular states and pathway activity; spatial measurements locate them relative to basement membrane, follicles or wounds. A trajectory inferred from transcriptomes is not a directly observed lineage, and an intermediate expression profile can reflect a transient response rather than a cell reversing fate. Combining molecular profiling with permanent labels, repeated imaging and appropriately timed perturbations distinguishes state similarity, ancestry and causal requirement. cell cycle state and injury-associated inflammation also need to be separated from a genuinely distinct renewing population.

Direct evidence for cell fate plasticity requires more than changing proliferation within an existing progenitor class. Pre-injury labelling of GATA6-expressing differentiated sebaceous-duct cells showed that their descendants could move into wounded interfollicular epidermis, regain basal renewing properties and produce a broader range of epidermal descendants. Long-term tracing and live observations supported dedifferentiation under the wound conditions. ^[ref] The key comparison is the fate of the same pre-labelled population in undamaged and damaged tissue. It is not sufficient to find Gata6-lineage fluorescence in a cell that has just begun expressing the marker after injury.

Several controls make such a claim persuasive. Verify the initial labelled cell identities before damage; allow the inducer to clear; exclude rare pre-labelled stem cells or reporter leakage; and demonstrate sustained renewal rather than transient basal residence. Imaging a labelled cell's transition and showing its later functional output is stronger than a static double-positive image. Exclude cell fusion and selective expansion of an unrecognized initial subpopulation where relevant. Perturbing a proposed reprogramming regulator should preserve source-cell viability, since losing the source is not the same as preventing its identity change.

The general conclusion is a flexible hierarchy of evidence. Label retention measures division history, clonal tracing measures descendant output, live-cell imaging measures event order and local interactions, and molecular profiling measures state. Together they support dynamic population self-renewal, site-dependent heterogeneity and injury-induced epithelial plasticity, while leaving normal maintenance distinct from regenerative potential. Claims about one mouse skin region or oesophagus should be tested independently before being generalized to every squamous tissue or to humans.

The Lgr5-expressing population is not always indispensable for short-term small-intestinal maintenance, but the unqualified claim of dispensability is misleading. Its contribution and its requirement depend on injury, timing, ablation efficiency and replacement by other cells. Three questions must be separated: whether Lgr5-expressing cells normally generate the epithelium, whether the Lgr5 receptor itself is necessary, and whether the tissue can replace the marker-defined cell state after its cells are removed.

In adult mouse small intestine, actively cycling crypt-base columnar cells include an Lgr5-expressing population that produces the major epithelial lineages. Inducible permanent lineage tracing from Lgr5-expressing cells generated long-lived epithelial contributions, demonstrating self-renewal and multilineage output. ^[ref] This identifies a functional stem-cell population under those conditions. It establishes normal contribution, however, not exclusive capacity or necessity. A lineage can supply most routine replacement while another population has latent capacity to take over after removal.

stem-cell ablation makes a different causal test. Engineering the human diphtheria toxin receptor into the Lgr5 locus allowed the marked population to be depleted. In the tested depletion regime, intestinal epithelial maintenance continued, and other cells could replace the missing stem-cell output; Bmi1-associated lineage experiments supported a reserve contribution. ^[ref] This result refutes a simple model in which immediate loss of all renewal must follow removal of the cycling Lgr5-positive compartment. It does not establish that no stem-like function is necessary, that the marker-defined state can never reappear, or that the same result holds indefinitely under every physiological challenge.

The relevant conclusion is about the experimental window and endpoint. Histologically recognizable crypts shortly after treatment are weaker evidence than sustained production of all lineages and preserved barrier or absorptive function. transit-amplifying cells and surviving differentiated cells can temporarily conceal a loss of long-term renewal. Conversely, if output continues beyond their expected persistence and previously unlabelled cells restore long-lived clones, genuine compensation is likely. Repeated observation should therefore quantify target-cell depletion, proliferation, crypt survival, lineage composition and the re-emergence of Lgr5 expression throughout treatment and recovery.

Interpretation also depends on whether reserve cells form a permanently distinct hierarchy or are recruited from existing descendants. Expression of a proposed reserve marker does not prove that it exclusively labels a quiescent stem-cell population. Drivers can label several states, marker expression can change after injury, and apparent populations can overlap. An ablation-induced contribution from a marker-labelled cell is evidence of regenerative potential in that labelled lineage; it need not reveal its normal developmental position. Labelling before injury and allowing the inducing reagent to clear are essential to avoid marking newly activated cells and mistaking them for the original source.

Committed descendants provide direct evidence for such flexibility. Enterocyte-lineage progenitors labelled through Alpi normally give short-lived absorptive descendants, yet after Lgr5-cell ablation they can produce persistent renewing lineages. ^[ref] This supports intestinal stem-cell plasticity rather than an absolutely one-way hierarchy. It should not be generalized to every mature cell on a villus: a progenitor already biased toward an absorptive fate and a terminally mature enterocyte are different starting populations. Clonal persistence and renewed multilineage output are stronger endpoints than transient expression of one stem-cell marker.

Further work connected recovery of the ablated intestinal stem-cell compartment to Ascl2-dependent dedifferentiation of absorptive and secretory precursors. ^[ref] Regenerative cells can initiate the returning programme before the full usual marker combination is restored. This makes stemness a context-dependent functional state as well as a useful marker-defined category. It also yields a testable limit to compensation: if a required reprogramming regulator is disabled, the presence of alternative surviving cells need not be sufficient for recovery. Targeted perturbation should be assessed alongside precursor survival so that failure to reprogramme is not confused with removal of the proposed source.

Dispensability changes markedly when another injury removes or disables the compensating population. Combining Lgr5-cell depletion with radiation-induced damage caused severe crypt loss and failure of robust regeneration, unlike the relatively preserved homeostasis after depletion alone. ^[ref] This demonstrates conditional necessity in the tested injury regime. It does not create a contradiction with compensation under ordinary maintenance: the two interventions alter different components of the regenerative system. Radiation affects multiple epithelial and niche populations, and its timing relative to ablation and recovery is therefore part of the explanation rather than a minor technical detail.

A factorial experiment makes that logic explicit. Compare control, Lgr5-cell depletion alone, radiation alone, and the combination, with matched ages, intestinal regions and observation times. Include receptor-negative animals receiving toxin, receptor-positive untreated controls, and independent confirmation of cell removal rather than relying solely on loss of the fluorescent reporter. Measure crypt number and clonogenic recovery, epithelial composition, apoptosis, inflammatory or stromal damage and longer-term function. Evidence that surviving untargeted cells generate new stem-cell clones is required to identify the replacement mechanism; normal tissue appearance does not identify the source.

The duration of targeting is also important. A lineage that first lacks the marker but later adopts Lgr5 expression may become sensitive to continued toxin treatment. A pulse of ablation followed by repopulation and sustained elimination of the re-emerging state consequently test different hypotheses. Escaper cells with incomplete receptor expression or recombination can expand after treatment and mimic recruitment from another lineage. Pre-existing lineage labels, spatial imaging and independent ablation strategies help distinguish rare survivors from true conversion, while acknowledging that every strategy can perturb the niche itself.

Finally, removing Lgr5-expressing cells is not equivalent to deleting the Lgr5 gene. The former removes entire cells, including their other molecular functions; the latter changes a receptor while leaving the cells and possible compensating pathways in place. Likewise, demonstrating organoid growth under supplied matrix and growth factors establishes competence under those conditions, not niche independence in vivo. Normal renewal, replacement capacity and receptor requirement must therefore remain separate conclusions.

The evidence supports a plastic, partly redundant renewal system: Lgr5-positive cells are major normal contributors, can be replaced under some maintenance conditions, and can become crucial when injury limits alternative sources. The scientifically useful claim is conditional dispensability of a defined population under a defined protocol, not general irrelevance of those cells or the stem-cell state they exemplify.

cell specification and morphogenesis can form a reciprocal causal loop: cell identity changes force production and tissue material properties, while the resulting geometry and loading alter signalling and differentiation. A molecular programme need not first assign every identity and then leave mechanics merely to implement a predetermined shape. Conversely, correlation between a new shape and a new marker does not establish that mechanical input instructed the fate. Developing intestine, avian skin and lung illustrate distinct mechanisms and experimental tests.

Cells generate forces through actomyosin contractility, transmit them through cell adhesion and the extracellular matrix, and respond through mechanotransduction. Changes in cell number, volume, polarity and matrix composition modify the mechanical problem. A deformed tissue can change ligand distribution, receptor exposure or intracellular signalling without any change in a ligand's production rate. One must therefore distinguish force generation, transmission and sensing. A contractility inhibitor can disrupt all three as well as cell division, so an abnormal organ alone is not a mechanistic explanation.

In the embryonic chick intestine, differentiation of surrounding smooth muscle changes the constraints on epithelial growth. Successive muscle layers restrict expansion in different directions, helping a growing epithelial–mesenchymal composite form ridges, zigzags and eventually villi through buckling. Muscle perturbations and artificial confinement support a causal role for physical boundary conditions rather than a purely epithelial prepattern. ^[ref] Here an initial differentiation event supplies a mechanical constraint that changes tissue shape. Equivalent external confinement is especially informative because it can replace a mechanical function without restoring all molecular products of the missing muscle.

The new geometry then feeds back into regional cell-state regulation. Epithelial Sonic hedgehog signals become concentrated beneath the curved villus tips, promoting mesenchymal BMP signalling; the resulting signals restrict the stem-cell-associated proliferative programme away from those tips. ^[ref] Thus spatially patterned signalling can arise from deformation of a source that was initially more uniform. The result concerns localization of a developmental stem-cell programme, not proof that every cell has irreversibly acquired one terminal identity. Comparing curved and flattened explants, measuring ligand production as well as local pathway activity, and changing pathway activity downstream of deformation distinguish altered geometry from a change in total ligand supply.

This chick mechanism should not be imposed unchanged on mouse intestine. Mesenchymal patterning and surface deformation have different temporal relationships in different species. The reusable principle is a feedback between differentiated supporting tissue, physical geometry and local signalling; the exact sequence is an empirical question. A useful control compares developmental stage and viability, since a tissue that has simply stopped developing may lack both folds and differentiation markers without either being the direct cause of the other.

Avian skin supplies a different route from mechanics to identity. Contractile dermal cells aggregate beneath the epidermis, mechanically compressing it. Experiments linked this dermal self-organization to activation of epidermal beta-catenin and initiation of the feather-follicle programme. Altering contractility or tissue mechanical conditions changed aggregate formation and pattern. ^[ref] This makes the feather placode an emergent epithelial–mesenchymal structure rather than requiring a complete molecular placode map to exist before any cell movement.

The important distinction is between beta-catenin's junctional and transcriptional roles. Redistribution of an adhesion-associated protein alone is not evidence for a fate change; nuclear activity and appropriate target expression, followed by placode development, provide a stronger chain. To test ordering, one can alter dermal force production while preserving epidermal viability, apply controlled compression, and perturb the epithelial response separately. Restoration of placode gene activity by a downstream intervention would place that activity after mechanical input, but would not necessarily rescue normal spacing. Molecular identity and periodic geometry are related outputs that can be experimentally separated.

In the developing lung, differences in how epithelial cells experience a common physical input help generate different differentiated states. live-cell imaging and perturbations showed that forces associated with fetal breathing movements and fluid displacement contribute to type I alveolar cell differentiation. A subset of progenitors protrudes toward mesenchyme through FGF10–FGFR2-dependent signalling; enrichment of apical myosin in those cells resists flattening and supports type II alveolar cell fate. ^[ref] Thus a local biochemical signal changes the cell's mechanical response, while loading helps determine whether it becomes thin and extended or retains the alternative morphology and differentiation programme.

A decisive lung experiment must therefore vary force and growth-factor signalling independently. If a receptor perturbation changes fate, it could act through competence, proliferation, survival or the physical protection from flattening. Cell tracking and early marker measurements distinguish transformation of an existing progenitor from selective loss of one population. Controlled loading, matched fluid and culture conditions, measurement of cell shape, and restoration of the relevant contractile response can separate these alternatives. Flattening alone is not a sufficient endpoint: the molecular and functional differentiation of the two alveolar classes must also be assessed.

These systems demonstrate three forms of coupling: differentiation changes a tissue constraint, geometry reorganizes an extracellular signal, and local signalling tunes a cell's response to a physical force. They need not all use the same sensor or transcription factor. Nor does mechanics replace biochemical information: the same deformation can elicit different responses in cells with different developmental competence. A fate programme can also alter adhesion or matrix properties and thereby change the mechanical input experienced by neighbouring cells, closing a mechanochemical feedback loop.

The strongest general strategy combines time-resolved live-cell imaging, fate or lineage reporters, a physical measurement and an independently targeted perturbation. Initial recoil after laser ablation reports load relative to resistance rather than an absolute force without a material model; substrate deformation and pressure measurements likewise need calibrated assumptions. Acute and tissue-restricted perturbations, sham procedures, independent genetic alleles and appropriate rescue help avoid generalized damage as the explanation. Fate and form are coupled through specific causal chains, but identifying a mechanically instructive step requires measuring both the physical intermediate and the resulting change in developmental state.

Mutual support is a major component of evolution, but its universal superiority to competition is neither a well-defined nor an established empirical claim. cooperation and conflict often explain different levels of the same biological system. The productive question is how support remains advantageous and stable when individuals or genetic elements can gain by contributing less, and how its ecological effects compare with the costs of competition in a specified setting.

In Mutual Aid: A Factor of Evolution by Peter Kropotkin, Peter Kropotkin emphasized animal associations and cooperative benefits that an exclusively combative picture of nature overlooks. ^[ref] This is a historically important corrective, not a controlled survey assigning a universal effect size to helping. Its descriptive examples generate hypotheses about survival and reproduction. They cannot establish how common cooperation is after correcting for which taxa, environments and conspicuous associations were observed. A claim about the entire organic world also exceeds a collection concerned principally with selected animal and human examples.

Definitions matter. mutualism supplies benefits to both participants, whereas evolutionary altruism imposes a direct fitness cost on an actor while benefiting another. cooperation can be mutually beneficial and require no sacrifice at all. Moreover, a short-term feeding cost can be part of a relationship that ultimately increases the donor's own lifetime reproductive success. The biological categories depend on the stated comparison and timescale, not on whether an action appears generous. Conversely, physical aggression is not required for competition: individuals can reduce one another's reproduction by consuming the same limiting resource without meeting.

Some support is explained by immediate complementary returns. Group defence or joint foraging can be individually advantageous when combined actions achieve something inaccessible alone. Field work on groupers and giant moray eels documented recruitment and nonrandom associations during hunting, consistent with gains from their complementary ways of accessing prey. ^[ref] The evidence concerns a particular interspecific association; it does not imply that all joint hunting is reciprocal altruism. To estimate the benefit rigorously, compare prey availability, effort, detection and success with and without the partner, accounting for selective association in especially profitable locations. Shared location alone would be weaker evidence than coordinated recruitment and complementary actions.

Support among relatives can evolve even when its actor pays a direct cost. In a simplified marginal accounting, let $C$ be the actor's loss, $B$ a recipient's gain, and $r$ the appropriate genetic association between them. An altruistic change is favoured when:inclusive fitness theory makes clear why effects on other individuals can influence selection on an actor. ^[ref] The coefficients and effects require specification: relatedness is not simply physical proximity, benefits must be measured in relevant reproductive units, and local demographic competition can change their net value. The expression is not a licence to label every action benefiting relatives adaptive without measuring costs and alternatives.

Experiments with siderophore production in Pseudomonas aeruginosa show why both association and competition matter. Manipulating relatedness and the scale of competition changed the success of cooperative production. ^[ref] High relatedness can make benefits to neighbours genetically valuable, but the same neighbours may compete for the same limited opportunities. Under sufficiently local competition, one relative's gain can replace another's rather than increase their collective contribution to future populations. Experimental mixing, replicated populations and controls for iron availability and bacterial growth help identify these mechanisms. The mere observation that producers grow better together than nonproducers would leave the evolutionary competition between the types unresolved.

Support among unrelated individuals has additional routes. by-product mutualism can supply immediate returns; partner fidelity feedback makes a persistent partner's improved condition beneficial to the actor; partner choice can redirect future opportunities toward good partners. Conditional withdrawal of benefits can limit exploitation. These mechanisms do not require all participants to value the group's success independently of their own reproduction. Their presence also does not eliminate evolutionary conflict over how the joint benefit is divided. A mutualism can remain beneficial to both participants while each favours a different exchange rate.

Even highly integrated societies contain competition over reproduction. In honeybees, workers can produce male eggs, and experiments showed discrimination against worker-laid eggs, supporting worker policing. ^[ref] Relatedness under multiple mating can contribute to the incentive to suppress worker reproduction, while effects on colony productivity also require consideration. The observed collective organization is therefore maintained partly by controlling opportunities for individual reproduction. Egg transfer experiments, controls for egg age and handling, and tests of recognition cues distinguish policing from generalized removal of damaged eggs. A peaceful-looking colony need not be free of conflict, and aggressive enforcement can help maintain collective functioning.

The same logic extends to integration across biological levels. A multicellular organism depends on coordinated cell activities, but somatic evolution can favour cell lineages that undermine organismal reproduction. A single-cell bottleneck, regulation of cell division and restrictions on reproductive lineages can reduce such conflicts. These are candidate mechanisms for alignment, not evidence that every part of the organism has always had identical interests. Major evolutionary transitions are important partly because cooperation becomes stable enough to form new reproductive units; suppression of lower-level competition can be part of that success rather than its conceptual opposite.

multilevel selection provides another accounting of these processes. Groups with more contributors can produce more descendants while noncontributors increase within mixed groups. Whether contribution spreads depends on the strength of both effects and on how groups form and reproduce. Simply demonstrating that a group performs well does not show that its costly contributing trait resists invasion. In appropriate models, inclusive-fitness and multilevel descriptions can track the same underlying change from different partitions, so changing vocabulary alone does not supply a different biological mechanism.

Assessing relative importance requires an explicit metric. One might compare effects on population persistence, individual lifetime reproduction, biomass production or the rate of a major transition. Those answers need not agree. A short interval of severe competition can determine the outcome despite long periods of cooperation; alternatively, cooperative defence can be essential during a rare environmental crisis. Support within a group can also improve that group's capacity to compete with others, transferring rather than eliminating antagonism. Neither the proportion of observations that look peaceful nor the number of visibly aggressive events measures total selective importance.

The defensible conclusion is that mutual support is pervasive and often indispensable, while evolutionary competition and conflict help explain when that support evolves, how it is regulated and where it breaks down. Mechanistic and comparative tests support this joint account more strongly than a universal ranking of support over struggle.

interaction frequency changes evolutionary incentives by altering immediate encounter benefits, learning opportunities and expected future returns. Its effect depends on who meets whom and what happens during an encounter; more interaction does not universally favour cooperation. Three quantities must be separated: encounter rate per unit time, probability of meeting the same partner again, and the frequency of a behavioural or genetic variant in the population. They can change independently.

Repeated encounters can support direct reciprocity. In a simple donation game, cooperation costs its actor $c$ and supplies a partner benefit $b$, with $b>c>0$. Suppose both players initially cooperate and use grim trigger, withdrawing cooperation permanently after a defection. With effective continuation factor $0\le \delta <1$, continued cooperation returns:whereas a one-time defector obtains $b$ and then receives zero thereafter against that rule. cooperation is a best response when:The repeated-game analysis shows how a cost now can be offset by future returns. It does not show that this rule must evolve, that it can invade from arbitrarily low frequency, or that animals calculate the inequality. Conditional strategies, initial assortment, implementation errors and monitoring determine the evolutionary outcome. The classic theoretical and computational work on repeated cooperation established the importance of future interaction without proving one universal mechanism in nature. ^[ref]

Elapsed time matters as well as the number of rounds. For illustration, let further encounters with the same partner occur at rate $\lambda$, while the relationship ends irreversibly at independent rate $\mu$. Under these competing exponential waiting times:Faster recurrence can then make benefits arrive before the relationship ends. Combining the two results gives $\lambda \ge \mu c/(b-c)$ under these restrictive assumptions. If frequent encounters are instead with new unrecognizable partners, they need not increase the continuation value of any one relationship. Higher population density can therefore increase encounter count while making partner-specific reciprocity less important.

Intraspecific food sharing in vampire bats supplies relevant evidence. Controlled fasting induced opportunities to donate food among familiar individuals of mixed relatedness; food received predicted food given more strongly than relatedness, and donors often initiated transfers. ^[ref] This supports a role for mutual direct benefits beyond a purely kin-based or harassment explanation. It does not establish one-for-one repayment after every donation. Stable social relationships can jointly cause both directions of help, so prospective manipulations of partner availability and ability to reciprocate would provide stronger tests of particular contingency rules. Repeated records within a dyad must not be counted as independent relationships, and opportunities to give must be distinguished from willingness.

Repeated interaction can also support cooperation between species. Cleaner fish can gain by consuming client mucus rather than the material whose removal benefits the client. In experiments using artificial clients, cleaners chose between preferred prawn and less preferred flake food; taking the preferred item caused plate removal or a brief chasing response in treatment groups. Cleaners subsequently changed their choices relative to controls. ^[ref] This tests a learning mechanism through which loss of access or an adverse response changes incentives. It does not directly measure a natural client's lifetime return or require a tit-for-tat rule. The controls against mere sequence-dependent preference change and the known initial food preference make the interpretation more precise than simply observing frequent cleaning.

partner choice adds a different effect of encounter structure. If clients can select among several cleaners, access to alternatives can reward good service through continued custom. A resident client with few alternatives may instead rely more on immediate control. Increasing the availability of partners can thus strengthen a biological market while reducing fidelity to any single partner. The net outcome depends on information about service, switching costs and competition for clients. Opportunity for future exchange matters, but partner switching and repeated bilateral reciprocity are not the same process.

Some mutualisms rely on physiological coupling rather than repeated behavioural encounters. In the legume–rhizobium system, experimentally preventing nitrogen fixation altered host support and bacterial reproductive outcomes, providing evidence for conditional partner sanctions. ^[ref] A persistent association can make a partner's contribution affect its own future resources without recognition of discrete past exchanges. The experiment should be interpreted through host physiology and bacterial reproduction, not as evidence that a plant consciously punishes. Controls for the manipulated gas environment and nodule condition distinguish reduced service from incidental treatment damage.

More frequent interactions can equally intensify antagonism. In a simple well-mixed predator–prey model, encounters can scale with the product of predator and prey densities. However, successful attacks cannot rise without limit when handling takes time. An illustrative Holling type II functional response is:with prey density $N$, encounter coefficient $a$ and handling time $h$. Increasing prey availability changes the value of search and prey choice, while saturation changes the benefit of further encounters. ^[ref] This is an ecological link to evolutionary incentives, not proof that all predators have the same response. Spatial refuge, prey switching and learning can alter it. Within species, frequent contests can instead increase injury, resource depletion or the value of stable dominance relationships.

The frequency of strategies supplies a further distinction. The return to aggression, cheating or a mating tactic depends on the alternatives encountered. A rare exploitative strategy may gain from common cooperative partners, but its advantage can decline as it depletes those opportunities. That is frequency-dependent selection, even if each individual has exactly the same number of encounters per day. By contrast, multiplying every payoff and cost by the same encounter rate may only speed selection without changing the favoured strategy. Effects on equilibrium require a change in relative returns, timing, partner composition or constraints.

Empirically, studies should manipulate encounter rate separately from identity, density and total resource availability, then measure survival or reproduction as well as immediate behaviour. Repeated partners can be maintained with fewer total meetings, or shuffled while holding meeting number fixed. Matched feeding and handling treatments help separate social experience from nutritional state and disturbance. Interaction frequency is therefore a mechanism-changing ecological variable, not a simple cooperation switch: its importance lies in how it reshapes future consequences, learning, assortment and the balance between mutual benefit and exploitation.

Stable social actions can reflect stable states or constraints, while flexible actions can follow an unchanged decision rule. The key distinction is between the rule, its inputs and the behaviour observed. A social decision rule maps perceived social information and internal state to an action. It need not involve conscious deliberation, and it need not change genetically whenever the action changes.

For example, an individual that helps when a cue exceeds a threshold follows a fixed conditional rule even if its daily actions vary. A simple social reaction norm is:where $S$ describes a specified social condition, $a_i$ baseline behaviour and $b_i$ responsiveness. Individuals can differ in either parameter. A stable difference in average actions does not establish that their slopes are zero; conversely, changing actions after a partner changes does not establish that the underlying rule has been relearned. Experiments should vary social conditions within individuals and compare both baselines and responses.

Selection favours responsiveness when information predicts changes in the relative benefits of actions. A partner's willingness to cooperate, current rank, resource availability or the need for a particular task can all alter those returns. Flexibility is less useful when inputs are unreliable, conditions change faster than an individual can respond, or alternative actions cannot be produced. The cost of behavioural flexibility can include sampling, neural processing and switching costs. A persistent environment may favour specialization, but environmental constancy alone is insufficient: social partners can still generate changing conditions.

Development and morphology can restrict the feasible alternatives. Body size, age, reproductive development and task-specific structures influence what an individual can do. Some specializations involve effectively irreversible developmental commitment, whereas others are reversible states. It is therefore necessary to measure capacity separately from willingness. An animal that remains in one role because it cannot perform another has a different explanation from an animal that could switch but currently gains more by staying. An individual's role can also be constrained by partner aggression or exclusion, so lack of observed switching need not reveal its preferred action.

age polyethism in honeybees illustrates both regularity and flexibility. Workers commonly change tasks with age, but rearing and transplant experiments showed that interactions with older workers alter behavioural development and its hormonal regulation. ^[ref] The pattern is thus not adequately described as an inflexible age clock. Changing the social composition can change when younger workers begin foraging. A convincing test distinguishes social information from nutrition, colony size, handling and brood demand, and measures task performance rather than relying only on age or a hormone concentration. The colony, not every observation of a worker within it, is often the independent experimental unit.

Social opportunity can also change an adult's behavioural state rapidly. In Astatotilapia burtoni, removal of a dominant male allowed subordinate males to adopt dominance-related behaviour and coloration within minutes, accompanied by a localized immediate-early gene response. Stable dominant and subordinate comparison groups helped distinguish new opportunity from simply performing dominance behaviour. ^[ref] The manipulation minimized disturbance to the focal fish. It did not establish that every component of reproductive physiology changed immediately: behaviour, gene expression and gonadal development operate on different timescales. A behaviourally fixed-looking subordinate can therefore possess latent flexibility that is expressed only when social constraints change.

Information acquisition and learning add another source of conditional behaviour. Repeated encounters can reveal partner quality, while experience can change the expected cost of aggression or cooperation. A winner-loser effect can reinforce existing social asymmetries, but prior contest outcomes are correlated with fighting ability unless experiments separate experience from pre-existing differences. Likewise, familiarity can alter interactions through recognition without requiring a complex representation of another's intentions. Testing partner-specific rules requires swapping partners or their cues while controlling location, experience and motivational state.

Stable individual differences can arise through inherited variation, persistent developmental experience or feedback between behaviour and state. An exploratory individual may gain information that makes further exploration advantageous; a dominant individual can gain resources that maintain its competitive advantage. Partners can adjust around these differences, producing social niche specialization. Once complementary roles are established, unilateral switching can be costly even if an alternative allocation would have been possible initially. Such stability need not imply a separate genetically fixed strategy for every role.

animal personality and behavioural repeatability describe patterns that require explanation rather than supplying one. Repeatability can be high because individuals repeatedly encounter different environments. A common garden experiment, repeated measurement across social contexts and controlled developmental histories help distinguish persistent environmental effects from inherited contributions. Even a heritable baseline can coexist with a highly flexible response slope. Conversely, averaging across opposite responses can hide substantial flexibility or falsely suggest that all individuals follow the same rule.

Social specialization should also be distinguished from efficiency. In ants, separately measuring task specialization and individual task performance showed that specialization did not generally predict greater efficiency. ^[ref] Stable allocation can therefore reflect task availability, differences in response thresholds or social organization rather than practice-driven superiority at the chosen task. Manipulating task demand and group composition, then measuring both the individual's response and colony output, is more informative than assuming that a frequently observed specialist is the best performer.

An experimental programme should consequently manipulate the information available and the payoffs or constraints separately. Change the number or identity of partners, task demand, social rank or encounter history; use sham disturbances and within-individual comparisons where appropriate. Reversing the treatment tests whether a change is reversible, while a delayed assay tests whether it persists through learning or developmental modification. Pedigrees or controlled breeding can estimate inherited variation without equating it with immutability. Response latency, errors and fitness consequences matter alongside the final action, because a rule can be flexible yet too slow or inaccurate to be advantageous.

Fixed and flexible social behaviour are outcomes of selection on conditional responses, information, state and feasible actions. Strong evidence identifies which component changed. Behavioural consistency alone establishes neither genetic determination nor adaptation, and behavioural change alone establishes neither learning nor a changed social rule.

An organism contains genetic elements and cell lineages whose transmission interests need not coincide. Conflict arises when increasing an element's representation in descendants or within the body differs from maximizing the organism's reproductive success. Shared residence does not guarantee shared evolutionary interests, but neither does every difference between genes constitute conflict. A convincing account identifies the competing lineages, their transmission routes, and the outcomes favoured by selection on each.

Ordinary Mendelian segregation helps align the interests of unlinked nuclear loci: most depend on the same successful organism and receive an approximately equal chance of entering its gametes. intragenomic conflict becomes possible when an element bypasses that shared route, biases transmission, replicates within the genome, or has a parent-specific effect on relatives. mutation supplies variation; genetic linkage can protect combinations involved in distortion, while recombination and selection at other loci can favour suppression. The fitness currency of a selfish genetic element is therefore its own propagation, not the apparent health or fecundity of its carrier.

meiotic drive provides a direct example. For a rare autosomal driver with transmission probabilities $t_f$ through heterozygous females and $t_m$ through heterozygous males, suppose, for illustration, that heterozygotes have the same viability factor $1-s$ in both sexes. With random mating and otherwise equal reproductive contributions, the rare-driver approximation is:A rare driver can increase when the multiplier exceeds one, despite a carrier cost. This accounting also shows why the sex in which drive occurs matters: with unbiased male transmission, $t_m=1/2$, not $t_m=t_f$. The approximation omits dominance in common genotypes, sex-specific fertility, linkage and suppressors; it is an explanatory invasion calculation rather than a quantitative model of every natural driver.

Asymmetric female meiosis creates one route to distortion because some products become eggs and others are discarded. In monkeyflowers, a centromere-associated allele showed a strong female transmission advantage, with male fertility costs that expose a conflict between transmission and organismal performance. ^[ref] Reciprocal crosses and appropriate genetic backgrounds are important: distorted counts among surviving offspring can result from embryonic viability differences rather than biased segregation. Direct observation or genotyping of gametes and early embryos helps identify the stage. centromere drive is consequently more specific than merely observing segregation distortion.

The evolutionary response can include drive suppressors at other loci. Suppression is favoured when restoring fair transmission or reducing fertility costs improves the propagation of the rest of the genome. Its effect depends on linkage: a modifier tightly associated with the driver can experience different selection from an unlinked modifier. Arms races can maintain polymorphism and alter reproductive mechanisms; conversely, a successful suppressor can conceal an earlier conflict. Present-day Mendelian ratios therefore do not establish that transmission has always been conflict-free.

Transposable elements exploit another route: copying or moving genetic material within a lineage. McClintock's maize experiments established mobile controlling elements through genetic and cytological observations, providing foundational evidence that genomic components need not occupy fixed positions. ^[ref] Mobility itself is not proof that every element is harmful or selfish in every context. The conflict hypothesis requires a replication or transmission advantage alongside consequences for other genomic functions. Host silencing through mechanisms such as RNA interference and DNA methylation can constrain propagation, while some formerly mobile sequences acquire useful regulatory functions. Current usefulness and historical origin are different claims.

genomic imprinting illustrates conflict through expression rather than transmission distortion. Maternally and paternally inherited copies can differ in their relatedness to individuals affected by resource allocation. If a mother's future offspring are less likely to share the current father, a paternally inherited copy in the current offspring can place relatively less weight on those future maternal offspring. kinship theory of genomic imprinting therefore predicts disagreement over resource extraction under specified mating and life-history assumptions. It does not predict opposed interests at every locus or in every mating system. ^[ref]

Targeted disruption of the mouse insulin-like growth factor 2 gene produced growth effects dependent on parental transmission, demonstrating parent-of-origin-dependent regulation of an important growth factor. ^[ref] This is mechanistic evidence for imprinting. It is not, by itself, proof that parental genetic conflict caused the evolutionary origin of that regulatory pattern. Testing that explanation requires comparisons among related species and mating systems, tissue-specific expression, effects on maternal and offspring fitness, and alternatives such as dosage regulation or coadaptation. The genetic copy's interest must not be confused with a father's conscious preference or with a universal claim that larger offspring are always better for one parent.

The broader category also includes conflicts between distinct genomes residing in one individual. cytoplasmic inheritance can make an organelle or microbial symbiont depend mainly on female transmission, whereas nuclear loci are ordinarily transmitted through both sexes. A maternally inherited symbiont can consequently spread through changes to reproduction that do not benefit every host. In Nasonia wasps, reciprocal crossing, chromosome observations and antibiotic removal connected cytoplasmic microorganisms to reproductive incompatibility and abnormal paternal chromosome condensation. ^[ref] This demonstrates a microbial cause of a barrier under those conditions, not necessarily the origin of every species difference. Antibiotic effects on host condition and other microbes require controls, and reintroduction experiments would strengthen microbial specificity.

somatic evolution is a related conflict at a different level. A mutant cell lineage can increase within a body by avoiding growth restraint or cell death while damaging the organism. Analyses of cancer genomes and normal somatic tissues identified patterns of selection on somatic mutations, showing that cellular success is an evolutionary quantity distinct from organismal success. ^[ref] Not all cell proliferation is selfish, and some detected mutations are passengers. Mutation-rate heterogeneity, tissue composition and sequencing ascertainment must be modelled before an excess of mutations is interpreted as selection.

Such conflicts have consequences beyond immediate harm. They can favour fair-segregation mechanisms, genome defence, regulated cell proliferation and restrictions on which cell lineages produce offspring. A single-cell bottleneck and germ-soma separation can increase alignment among cellular descendants or limit transmission of within-body competitors, although neither prevents new mutations. Conflict can also affect fertility, sex ratios, reproductive isolation and genome organization. Claims that it produced a particular major evolutionary change require historical and comparative evidence in addition to a plausible selective argument.

Organismal integration is partly an evolved outcome of controlling conflict, not evidence that conflict is absent. The strongest explanations connect a specific transmission asymmetry to an element's advantage, a measurable cost or opposed interest, and an experimentally or comparatively supported response.

Parents share an interest in successful offspring but can differ over mating, fertilization, resource allocation and the costs of care. Resolution usually means a stable outcome under those competing incentives, not disappearance of the underlying conflict. sexual conflict is defined by differences in the outcomes favoured by selection, and can occur without overt aggression. It should be distinguished from parent-offspring conflict, which compares parental and offspring interests rather than those of the two parents.

A fundamental source is the unequal return from additional mating or additional investment. gamete production, gestation, lactation and care can impose different constraints on the sexes, but those constraints do not dictate one universal sex role. What matters is how changing an action affects lifetime reproductive success, including future opportunities. A parent can favour more current investment by its partner while preferring to reserve its own resources for later reproduction. Conversely, both may benefit from increased joint effort when offspring survival requires contributions that neither can supply alone.

The conflict over mating begins before two animals become the parents of a particular brood. The mating rate that maximizes one sex's lifetime return can differ from that favoured by the other. Courtship persistence, resistance and ejaculate effects can then evolve through interlocus sexual conflict. In Drosophila melanogaster, comparisons involving males deficient in sex peptide connected a seminal component to costs of mating for females under the tested mating conditions. ^[ref] The result is stronger than a correlation between mating frequency and shortened life, but it does not mean that the component is always harmful: effects depend on exposure, nutrition, remating and the offspring produced. Genetic-background controls and lifetime offspring counts are necessary when separating altered mating behaviour from direct physiological effects.

antagonistic coevolution can generate matching manipulative and resistant traits. Water-strider research combined experimental mating outcomes with comparative analyses of male grasping and female anti-grasping structures across species. ^[ref] Such evidence supports an arms-race interpretation, while retaining alternatives: correlated structures can reflect shared ancestry, ecology or mechanical compatibility. Demonstrating opposed effects on male and female lifetime fitness is more informative than labelling every elaborate genital structure a weapon. evolution can reduce the current effect of a manipulation without removing the possibility of further counteradaptation.

After mating, the identity and number of offspring produced create further disagreements. A male can benefit from fertilizing a greater share of a brood, while a female's favoured allocation can depend on offspring quality, compatibility and future reproduction. Conflict over remating and fertilization should not be equated with conscious knowledge of parentage. paternity certainty can influence the return to paternal care only through available information and feasible alternatives; population-level extra-pair paternity is not itself evidence that an individual detects a particular offspring's parentage. A male may still benefit from caring for a mixed brood when withholding care would sacrifice his own offspring as well.

The logic of conflict over biparental care can be expressed with female effort $x$ and male effort $y$. Let offspring benefit be $B(x+y)$, with diminishing marginal returns, and let parental costs be $C_f(x)$ and $C_m(y)$. In a simple model, each parent values the same offspring benefit but pays only its own direct cost:At an interior best response with positive effort, each balances $B^{\prime }$ against its own marginal cost. The outcome need not maximize the unweighted sum of the two payoffs, because that sum counts the shared offspring benefit twice. This is a transparent model of opposing incentives, not a complete prediction of care in a species. Different parentage, offspring dependence, survival and future mating opportunities alter both benefit and cost functions.

When $B^{\prime \prime }<0$ and $C_f^{\prime \prime }>0$, the female's local response to an imposed change in male effort satisfies:Thus this particular model predicts partial compensation in parental care: reducing male effort increases female effort, but does not fully replace the lost care. Full compensation would give the reducing parent no immediate offspring penalty and could destabilize contributions. The slope is not a universal law, however. It assumes fixed information and benefit functions and does not apply unchanged when a partner's actions provide new information about brood demand.

Experiments on great tits used extra begging-call playback during one parent's nest visits to increase that parent's provisioning. The chicks' own begging did not increase, while the other parent also increased its effort. ^[ref] Such matching can occur when partner activity reveals that offspring demand or expected returns are higher than previously estimated. It therefore need not refute all conflict over care; it challenges the assumption that each parent treats the partner's effort only as a substitutable input. Experiments should manipulate offspring signals and partner actions separately, measure actual food delivered rather than visit counts alone, and compare immediate responses with later care and reproductive consequences. Handicapping one parent can also change brood condition, so an apparent negotiation response may otherwise combine several causal pathways.

parental negotiation can stabilize care through conditional adjustment, coordination and repeated interaction. Desertion opportunities, ecological dependence and availability of replacement partners can constrain unilateral withdrawal. Such constraints can align behaviour without aligning every preferred outcome. The observed result may involve alternating visits, task specialization or unequal contributions; equal time is not a necessary prediction. Conflict can also be reduced through sex-specific expression of a shared trait when its sex-specific optima differ, which concerns intralocus sexual conflict rather than manipulation between interacting loci.

Pregnancy extends the problem to allocation through offspring tissues. Under some patterns of relatedness, maternally and paternally inherited copies in offspring can favour different levels of maternal resource extraction. kinship theory of genomic imprinting concerns those copy-specific interests, not a claim that every father always favours larger offspring or every mother smaller ones. ^[ref] Maternal survival, future offspring and mating structure determine the relevant inclusive-fitness effects. An expressed physiological compromise can therefore coexist with genetic conflict over its regulation.

Conflict is resolved through the evolution of resistance, constraints, conditional care and negotiated equilibria, often with continuing costs. Establishing a mechanism requires comparing the parties' reproductive consequences and experimentally separating information, partner capacity and offspring demand; merely observing harmonious care or aggressive mating is insufficient.

Animal signals are diverse because their functions, receivers, transmission environments and evolutionary incentives differ. Similar functions can also be achieved by different historical and developmental routes. An animal signal is shaped by selection for its effects on a receiver; an informative incidental feature can instead be a biological cue. Merely finding that another animal responds to a feature does not establish why it evolved. Diversity includes differences among signal modalities, differences among species using the same modality, geographical variation, and variation among individuals or contexts within a population.

Different problems demand different information. mate choice can favour information about identity, condition or compatibility; contests can favour assessment of fighting ability or motivation; offspring solicitation concerns need and likely returns to provisioning. Alarm communication can coordinate responses to danger, whereas recruitment can direct partners to resources. A single individual may therefore use several signals because the appropriate receiver and useful message change across activities. The efficiency of a signal is defined relative to a receiver's response and its consequences, not simply the amount of information that a human observer can extract.

Physical transmission strongly constrains signal modality. Light can provide fine spatial patterning but depends on illumination and line of sight. Sounds can travel around obstacles, yet propagation, reverberation and background noise alter what reaches a receiver. Chemical signals can persist after a signaller leaves, which is useful for some functions and poorly suited to rapid reversal in others. Substrate vibration and electrical stimulation provide additional routes where anatomy and environment permit them. The relevant comparison is the complete communication system: production, transmission, detection and action. A brightly coloured patch is ineffective if it is hidden by posture or cannot be distinguished by the intended receiver.

The environment can consequently drive divergence in signals and sensory systems. In cichlids, variation in the light environment was associated with differences in visual pigments, male colours and female preferences, supporting a role for sensory drive in divergence. ^[ref] The inference requires evidence for each link; a correlation between water colour and male colour alone would not establish evolved receiver tuning or its effect on mating. Reciprocal illumination treatments, receptor measurements and preference tests help distinguish sensory mechanisms from population history. Actual mating and gene flow provide further evidence about consequences beyond laboratory attraction.

Background interference also changes which signals work. great tits in noisier urban locations sang with higher minimum frequencies in a classic field comparison. ^[ref] This illustrates a plausible response to acoustic masking, but the association by itself does not distinguish immediate adjustment, developmental learning, habitat sorting and inherited change. Playback of controlled noise to the same birds, repeated recordings, and matched quiet sites would separate some of these explanations. Signal amplitude, distance to the microphone and recording conditions also require control. Diverse responses to the same disturbance can reflect different receiver sensitivities or different costs of modifying a signal.

Selection by intended receivers can oppose selection by unintended ones. Male guppy colour patterns are associated with mating advantages but can also increase detection by predators. Experimental changes in predation and visual background altered colour-pattern evolution, showing that ecological conditions can change the balance rather than imposing a single universally optimal ornament. ^[ref] The intended receiver and predator may have different colour vision, so conspicuousness should be modelled and tested separately for each. Increased mating attraction does not establish a net lifetime advantage if survival is substantially reduced. signal eavesdropping thus helps explain why closely related populations can evolve different compromises.

Receiver psychology supplies further routes to diversity. A preference may originate in another task and subsequently be exploited by a new display, a process of sensory exploitation. Alternatively, a signal and preference may coevolve through sexual selection. These hypotheses have different historical predictions, requiring independent phylogenetic evidence about when the preference and display arose. Learned recognition can produce local conventions, while social learning can maintain differences without an initial genetic difference in the signal. A novel signal can be favoured when familiar ones lose effectiveness, but novelty is not always beneficial: unfamiliar warnings or mate signals may instead be ignored or mistaken for something else.

signal honesty also permits several designs. Reliability can follow aligned interests, physical links between production and an advertised property, or differences in the costs and benefits of expression. An index signal is constrained by its production mechanism; a condition-dependent signal may be easier for some individuals to sustain. In costly signalling, an expense is relevant when it changes incentives to exaggerate. Showing that a display consumes energy is not sufficient: an equally expensive display could still be advantageous to an unreliable signaller. Tests should manipulate both the signal and the advertised property, then measure receiver responses and the resulting fitness effects. Different information problems need not share the same honesty mechanism.

Conflict can maintain deceptive alternatives alongside informative signals. fork-tailed drongos use varied alarm mimicry to obtain food from other foragers, and changes in signal type can restore responses when targets become less responsive to repetition. ^[ref] This demonstrates how receiver experience can favour behavioural variation rather than one permanently effective deception. To separate signal-specific learning from declining general responsiveness, experiments need controlled playback order, repeated genuine alarms and measurements of target state. Deception also depends on an otherwise useful response remaining sufficiently common; a response that never benefits its receiver is unlikely to be maintained indefinitely.

Finally, developmental mechanisms and evolutionary history limit the available solutions. Related species inherit similar production structures and sensory systems, so some resemblance reflects biological homology. Different lineages may reach similar functional outcomes through convergent evolution. multimodal signalling can supply redundancy under variable conditions or combine different information, but additional components also carry production and detection costs. Comparisons should therefore measure both performance and history, account for phylogenetic nonindependence, and distinguish within-individual flexibility from inherited divergence.

Signal diversity is the outcome of several interacting filters: what benefits receivers, what signallers gain, what travels and can be perceived, and what inherited mechanisms can produce. No single explanation such as mate attraction, environmental adaptation or honesty is expected to account for every component of that diversity.

Genetic data provide relationships, demographic histories and mechanistic hypotheses that can be tested against fossil anatomy and dates; they do not make the synapsid fossil record redundant. Most deep synapsid evolution is much older than recoverable DNA. Molecular biology therefore often contributes indirectly through living descendants, calibrated models and developmental experiments, while exceptional younger fossils yield Ancient DNA or proteins themselves.

A major contribution is an independent framework for relationships. Superficially similar “ungulates” or insectivore-like mammals need not form natural groups. Molecular analyses recovered large placental groupings such as Afrotheria and Laurasiatheria, each containing independently specialized ecological forms. ^[ref] That result changes interpretation of the fossil record: similarities in feeding or locomotion must be tested for convergent evolution, and extinct taxa must be placed using broader character combinations. Living DNA does not itself score the missing anatomy of an extinct animal, but it changes which ancestral reconstructions and comparisons are plausible.

Whale origins show a particularly clear exchange between molecules and fossils. Molecular evidence placed Cetacea within the artiodactyl radiation and allied it with hippopotamuses, challenging interpretations driven strongly by dental resemblance to mesonychians. Analyses of independent retrotransposon insertion loci supported the whale–hippopotamus relationship. ^[ref] Subsequently described early whale limb skeletons supplied artiodactyl ankle characters consistent with that relationship. ^{[ref] [ref]} This is stronger than agreement among several reanalyses of the same tooth characters because the anatomical and molecular character systems have different sources of error. A living hippopotamus remains a relative, not a whale ancestor.

Rare genomic changes can be particularly useful, but no character type is infallible. A retrotransposon insertion requires orthology and correct reconstruction of insertion presence or absence. Ancestral polymorphism and incomplete lineage sorting can produce discordance among loci; hybridization can produce a genuinely non-tree-like ancestry. The same concerns apply to nucleotide data: one gene tree, especially a mitochondrial genealogy, need not equal a species tree. Broad genomic sampling and explicit models are therefore preferable to declaring a relationship settled by a single attractive marker.

Molecular clocks add time to the framework. Under the simplest two-lineage model, expected sequence divergence is approximately:with substitution rate $r$ per site per unit time and divergence time $t$. The sequences alone constrain a rate–time product, not both independently. Fossil calibration, other external time information and assumptions about rate variation are needed. fossil age, stem-versus-crown placement and the difference between a minimum occurrence and an origin are consequently central molecular inputs, rather than inconvenient qualifications added afterwards.

The timing of placental diversification illustrates the consequences. Some combined anatomical and molecular reconstructions emphasized a major post-Cretaceous–Palaeogene boundary radiation, whereas calibrated phylogenomic timelines support pre-boundary divergences for important deep placental lineages. ^{[ref] [ref]} These results need not disagree about every aspect of the event: separation of deep lineages, diversification within orders and expansion in ecological disparity can occur at different times. fossils identified only as Eutheria do not automatically establish a Cretaceous crown placental order. Conversely, failure to sample an earlier crown member is not automatically a hard maximum on its age.

The resulting ghost lineages formulate useful fossil predictions. A molecular hypothesis placing two living branches apart before their oldest known fossils predicts unsampled history, potentially in different regions or ecological settings. Appropriate tests compare collecting intensity and preservation, search suitable deposits and assess whether putative older members possess the required characters. Calibration sensitivity should be reported: changing maximum-age assumptions, rates or fossil placement can move the estimate. A date inferred using a particular fossil as its calibration is not an independent molecular confirmation of that same fossil's age.

Genomic and developmental work can also test a proposed transformation of anatomy. The mammalian middle ear incorporates bones homologous to components of the ancestral jaw apparatus. Living mammalian development retains a connection involving Meckel's cartilage, and experimental study of its breakdown, including interference with cartilage-resorbing cells, links developmental removal to the transition from a jaw-connected to a detached ear apparatus. ^[ref] These experiments offer a mechanism consistent with fossil intermediates. They do not show that every fossil lineage used an identical gene-regulatory change or that evolution simply repeated the entire embryonic sequence. To test a specific mechanism, compare matched developmental stages and relevant tissues and distinguish a targeted persistence of the connection from generalized developmental delay.

Younger fossils sometimes allow a more direct molecular test. collagen sequences from Toxodon and Macrauchenia supported affinities with the perissodactyl side of the placental tree, clarifying the relationships of South American ungulate-like mammals for which convergent anatomy had been problematic. ^[ref] Palaeoproteomics recovers amino-acid information rather than a complete genome, so the number and independence of informative positions differ from DNA studies. Authentication, contamination checks and comparison with plausible alternatives remain necessary, and the result should not be generalized automatically to every extinct South American ungulate lineage.

Ancient mammoth genomes extend direct inference into the Early Pleistocene. More-than-million-year-old DNA revealed distinct mammoth lineages and supported hybrid ancestry in the history of Columbian mammoths. ^[ref] This helps test whether a morphology-based fossil grouping corresponds to one ancestry lineage, reveals population processes invisible in an isolated bone, and connects some adaptive changes to dated specimens. Preservation in a favourable environment does not imply that DNA can be recovered from Permian or Mesozoic synapsids. The deep fossil record must therefore remain accessible chiefly through anatomy, chemistry and comparative models.

Finally, genetic data help separate the origin of an underlying pathway from the origin of a visible trait. A conserved developmental gene can be older than a particular ear structure or locomotor specialization; a gene duplication or regulatory change is not, by itself, a dated fossil of a phenotype. Functional experiments establish what a mechanism can do in a living context. fossil distribution and comparative anatomy establish when and where the corresponding construction is actually evidenced. Total-evidence phylogenetics can combine these sources, but must retain their different uncertainties and avoid counting correlated characters as independent confirmations.

“missing link” is useful when it names a specific, testable gap in an evolutionary reconstruction; it is misleading when it implies a single required intermediate in a linear chain. Modern phylogenetics predicts branching relationships, shared derived characters and sequences of anatomical change. A transitional fossil can preserve a mosaic of ancestral and derived features without being the direct ancestor of any named species. The surviving record samples relatives of ancestors far more readily than it identifies particular parent-to-descendant chains.

A helpful prediction therefore specifies what is missing. One might predict a therian relative that combines derived molar characters with an ancestral reproductive or pelvic condition, or an early aquatic placental retaining terrestrial weight-bearing limbs. The fossil should occur in an independently defensible time interval and be evaluated in a broad character analysis. Naming an unusual fossil after seeing it is weaker than showing that it tests an earlier prediction. Conversely, incompleteness of the fossil record is expected from preservation and sampling; absence alone does not falsify common descent unless suitable rocks, preservation and collecting effort make the absence informative.

Theria includes living marsupials and placental mammals and their common ancestry. Metatheria and Eutheria are broader lineages on either side of that split, whereas the living marsupial and placental crown groups are more restricted. These definitions are essential when a fossil is called a “link” to placental mammals. Possession of some eutherian characters does not establish that it belonged inside the crown group of living placental orders, much less that it was the ancestor of all of them.

Eomaia scansoria, from the Early Cretaceous of China at approximately 125 million years ago, illustrates the distinction. Its skeleton extended knowledge of early eutherian anatomy and combined derived characters used in phylogenetic placement with an ancestral overall construction, including epipubic bones. Limb and foot anatomy also supported climbing capabilities. ^[ref] It is informative as a commonly reconstructed stem eutherian, rather than as a complete miniature version of a modern placental. The retained epipubic bones are particularly valuable because they show why dental and cranial affinities must not be converted automatically into an inference of fully modern reproductive physiology. Soft-tissue details and exact gestation cannot simply be read from the skeleton.

A second example shows why a “missing-link” label must remain revisable. Sinodelphys was initially interpreted as an Early Cretaceous metatherian, apparently helping sample the early marsupial side of the split. Description of the more anatomically informative Ambolestes zhoui and revised comparisons instead supported a eutherian placement for Sinodelphys. ^[ref] The change did not make the fossil worthless or refute evolution. It changed which hypothesis its characters support, illustrating the role of missing anatomy, character homology and taxon sampling. Treating the initial identification as a permanent certificate of ancestry would make the metaphor obstruct rather than assist testing.

Transitions among established placental lineages supply more complete examples. Early whales test the prediction that an aquatic mammal lineage descended from terrestrial relatives. Pakicetids combine diagnostic whale ear-region anatomy with a substantially terrestrial skeleton. Protocetid ankle bones preserve the double-pulley astragalus characteristic of artiodactyl ancestry, even while the locomotor system was becoming increasingly aquatic. ^{[ref] [ref]} The significance lies in the combination: a whale need not already have every feature of a living dolphin, and the ankle provides an independent character system against which a relationship based on teeth can be tested.

Later basilosaurids were fully aquatic but retained very small hind limbs incapable of supporting terrestrial walking. ^[ref] Such remains test gradual modification of structures and changing functions, rather than requiring an organism halfway between two modern endpoints in every trait. A reduced structure is not necessarily functionless, and a named basilosaurid is not thereby demonstrated to be the direct ancestor of all living whales. The sequence is assembled from branching relatives and independently dated occurrences, with uncertainty about particular soft tissues and behaviours.

Pezosiren portelli provides a valuable independent transition in Sirenia. The Eocene Jamaican skeleton retains four substantial limbs, a multivertebral sacrum and a strong sacroiliac connection consistent with supporting the body on land, alongside aquatic specializations. ^[ref] It therefore contradicts the assumption that all members of the sea-cow lineage always possessed the obligately aquatic, hindlimb-reduced condition of living manatees and dugongs. Similarity to early whales in the broad transition to water does not make sirenians whale ancestors: they represent separate placental branches. Comparing the transitions helps distinguish common functional demands from common descent of the entire specialized body plan.

Marsupial history adds a biogeographic use of transitional character combinations. Early Eocene Australian Djarthia preserves ankle evidence informative about the early history of Australidelphia, which includes Australian marsupial groups and the South American Microbiotheria. ^[ref] It helps test relationships and former distributions without identifying one fossil as the unique ancestor that physically crossed a particular route. An early sampled occurrence sets a minimum age and a location, not necessarily an evolutionary birthplace. This is a more precise application of the useful intuition behind a geographical “missing link”.

These cases test several parts of evolutionary reconstruction: nested anatomical relationships, the timing of innovations, correlated changes in function and the geography of lineages. Strong assessment combines multiple structures rather than choosing a single convenient similarity, compares alternative placements, and treats geological age separately from anatomical inference. An apparent contradiction should first be checked for dating error, mistaken homology or sampling bias; alternatives must themselves make testable predictions rather than becoming unlimited excuses.