Methods & Confidence

BRACKETED The Extant Phylogenetic Bracket

Feathers, bones and eggshells fossilise. Lungs, air sacs, muscles and behaviour mostly don't. So how does anyone say anything about the soft parts of an animal that died 125 million years ago?

The answer is a piece of logic called the Extant Phylogenetic Bracket, and it is the single most important method on this site. Every extinct dinosaur sits on the evolutionary tree between two groups that are still alive: crocodilians on one side and birds on the other. Those two are the nearest living relatives we have. If a trait is present in both living groups, the most economical explanation is that the common ancestor had it too — which means the extinct animals bracketed between them almost certainly had it as well. That inference is strong, and it is what a BRACKETED tag on this site means.

The method is just as useful for what it refuses to say. When a trait appears in only one of the two brackets — in birds but not crocodilians, or the reverse — the inference is equivocal. There is no free pass. Something else has to carry the argument, usually a hard feature preserved in the fossil itself.

DIRECT A worked example: air sacs, and why the bracket alone isn't enough

Living birds breathe with a flow-through system: air is pumped through the lungs in one direction by a set of air sacs that reach into hollow spaces in the bones. Crocodilians also move air unidirectionally — but they have no air sacs and no pneumatic postcranial skeleton. So on air sacs specifically, the bracket splits. Birds yes, crocodilians no. Equivocal.

What closes the gap is something you can hold in your hand. Non-avian theropods preserve postcranial skeletal pneumaticity: the openings and internal chambers that, in living birds, are made by invading cervical and abdominal air sacs. The pneumatic bone is DIRECT evidence. The soft-tissue air sacs it implies are BRACKETED from the osteological correlate in living birds. Two tiers, stacked, doing different jobs — and the site labels them separately rather than collapsing them into one confident sentence.

INFERRED Melanosome colour — and the colours it cannot see

Fossil feathers can preserve melanosomes, the microscopic pigment-bearing organelles that give living feathers their blacks, greys, browns and rufous tones. Melanosome shape and packing density track colour fairly tightly in living birds, so the method compares the geometry and density of fossil melanosomes against a training set of extant-bird feathers, using quadratic discriminant analysis to assign a likely hue.

It has produced some of the most striking results in the field: a near-complete life reconstruction of Anchiornis (grey body, black-and-white banded wings, reddish-brown crest), countershading and a banded tail in Sinosauropteryx, and glossy iridescent black plumage in Microraptor.

Now the honest part. The method has real blind spots, and they matter:

• Structural colour is largely invisible. Iridescence and other structural colours depend on nanostructure spacing, which is generally not preserved.
• Whole pigment classes leave no trace. Carotenoid- and psittacofulvin-based colours produce no melanosome signal at all — so reds and yellows built from those pigments are simply not detectable. This is a direct caution for the Volume I bridge: parrot reds and yellows come from psittacofulvins, and melanosome science would be blind to them.
• Some specialists dispute the rigour of inferring hue from melanosome geometry alone.

So colour reconstructions on this site are presented as hypotheses with confidence bands, not as photographs. A grey-and-black Anchiornis is a well-supported reading of preserved evidence; it is not the same kind of fact as the bones.

How we communicate uncertainty — the five tiers

Every claim in Volume II carries one of five badges. The badge is not a measure of how interesting the claim is; it tells you what kind of evidence sits behind it, and therefore what you're entitled to conclude from it.

• DIRECT Preserved in the specimen. Somebody can point at it in the rock — Citipati sitting on its clutch, quill knobs on a Velociraptor ulna, filaments on Yutyrannus. You may treat this as observed. It is the strongest thing on the site.
• INFERRED An analytical result read out of preserved material. Melanosome colour reconstruction is the type case. You may treat this as a well-supported hypothesis with a stated method and known limits — not as a direct observation of the trait itself.
• BRACKETED Inferred from the Extant Phylogenetic Bracket. The trait is present in both crocodilians and birds, or is implied by a preserved osteological correlate. You may conclude the animal very probably had it. You may not conclude anyone has seen it in a fossil.
• CONTESTED Specialists actively disagree in print. You may conclude that the question is open — and that anyone telling you it's settled is ahead of the literature.
• SPECULATIVE Plausible, but lacking direct fossil support. Reasonable soft-tissue or behavioural reconstruction. You may enjoy it. You may not cite it as fact.

Two editorial rules keep the tiers honest. First, a claim without a resolvable primary citation defaults to a visible unverified state rather than being published as fact. Second, figures that trace to reputable secondary reporting rather than a primary table — the Yutyrannus mass estimate of roughly 1,414 kg is the standing example — ship tagged reported, verify against primary rather than quietly hardening into a number.

Open data & licences

Which sources Deep Time can reuse, what each licence requires, and the two that look open but are not.