Methods
Evidence tiers
Every record carries one of five tiers. T1 is a direct record of the thing itself. T2 is a documented account of it. T3 is an inference with a stated method. T4 is a reconstruction built from cited inputs. T5 is a speculative rendering. Tiers T3 to T5 always name a method page, list what they were derived from, and carry a confidence statement written by the person who made them.
Time
Years are stored as signed calendar years and displayed in Ma, ka, BCE or CE. Each record has a start, an end, a precision and, where the precision is coarser than a year, an uncertainty in years. On the timeline a record is fully visible when its range overlaps the window and fades across its uncertainty margin outside it.
Location
Locations are generalised at build time. See the ethics page for the rules. Generalised records are drawn at the centre of an H3 hexagonal cell (resolution 5 for about 10 km, resolution 4 for about 50 km) with a halo the size of the cell.
Reconstructed globes
For time steps before 5 million years ago, record positions and coastlines are reconstructed with the GPlates Web Service at build time using the rotation model named below. Positions are fetched once per step and stored as static files, so nothing depends on the service when you use the site.
Rotation model and data services
Rotation model: MULLER2022, via the GPlates Web Service, fetched on 2026-09-20 for time steps at 300, 200, 150, 75, 30 Ma. Coastlines are simplified for the web. Records are reconstructed from their public (generalised) positions, never from true coordinates.
Basemap: Natural Earth land and coastline vectors (public domain). Hexagonal cells: H3. Data sources by record: Ainmean-Àite na h-Alba; Chronos Terra Foundation; Chronos Terra field recordings; Chronos Terra field survey; Creswell Crags Museum; Digitised library; Glasgow City Council; Heritage group; Historical accounts; Holding archive; Lyme Regis Museum; NatureScot; Royal Ontario Museum; Saving Wildcats; Tobar an Dualchais; Wikimedia Commons; xeno-canto. See Sources.
Reconstruction methods
Every T3 to T5 record names one of these pages. Each page states inputs, what the method can and cannot tell you, and how confidence is expressed.
Comparative method (linguistic reconstruction) T4
Used by 1 record.
What the method does
Cognate words in related living languages are aligned sound by sound. Regular correspondences between them point to an ancestral form. That form is written with an asterisk to mark it as reconstructed. If a pronunciation is synthesised, it follows the written form and standard assumptions about the sound system.
Inputs
- Attested words in living or well-documented languages (the T1 or T2 records the reconstruction cites).
- Published etymological dictionaries for the family.
What it can and cannot tell you
It can propose the consonant and vowel skeleton of a word and place it roughly in time. It cannot recover accent, intonation or the voice of any speaker. No one has heard the word; the audio is a model.
Confidence
Stated per record. Usually good for consonants, weaker for vowels, none for exact pronunciation.
Crest resonance modelling T4
Used by 1 record.
What the method does
The nasal passages inside a lambeosaurine crest are traced from a scan of the fossil skull and treated as a resonating tube. The tube's length and shape set which frequencies it favours. Those frequencies are then synthesised so a listener can hear the model.
Inputs
- A cited fossil specimen with a preserved crest (the T1 record the reconstruction is derived from).
- Published airway measurements or a scan of the crest.
- A stated acoustic model (a simple tube model in this archive).
What it can and cannot tell you
It can suggest the fundamental and first few harmonics the crest would have reinforced. It cannot tell you the animal's timbre, loudness, rhythm, or whether it called at all. Soft tissue, vocal organs and behaviour are not preserved, so the output is a model of an instrument, not a recording of a voice.
Confidence
Moderate for the resonant frequencies (about 20 percent either way). None for timbre, rhythm and loudness.
Reference
Diegert, C. F. and Williamson, T. E. (1998). A digital acoustic model of the lambeosaurine hadrosaur Parasaurolophus tubicen. Journal of Vertebrate Paleontology 18(3).
Soundscape composition T5
Used by 1 record.
What the method does
A composer assembles modern recorded sounds (wind, water, insects) into a scene guided by what is known of a past environment. Nothing in the result is a reconstruction of a specific extinct organism.
Inputs
- A cited T1 record for the place or fossil assemblage the soundscape imagines.
- Published descriptions of the environment.
- Modern analogue recordings with their own rights and attribution.
What it can and cannot tell you
It can help a listener imagine a place. It carries no evidence about what the past sounded like. That is why it is labelled T5 Speculative, shown in outline only on the map, and hidden until the tier switch is on.
Confidence
None as evidence. It is an artistic rendering built from cited analogues.
Stratigraphic inference T3
Used by 1 record.
What the method does
A layer with no direct date is placed in time by its position in a described profile and by published dates for the same kind of layer in the region. The result is a range, never a point.
Inputs
- A described profile (the T1 record the inference is derived from).
- Published regional sequences, for example deglaciation chronologies.
What it can and cannot tell you
It can bracket a layer between known events. It cannot give a calendar date for the layer itself.
Confidence
Stated per record. Typically good at the scale of thousands of years and no better.