The Science That Destroys Its Evidence
Excavation is the only experiment that can never be rerun: to read a site you must take it apart, and the layers, once lifted, are gone. Archaeology's answer to this brutal fact is total recording — the site is destroyed, but it is destroyed into an archive of drawings, photographs, coordinates, and context sheets from which the sequence can be rebuilt on paper forever.
Before the First Trowel
Digging is the last resort, not the first move. A project begins with survey:
- Desk-based assessment — maps, archives, previous finds, aerial photos.
- Field walking — systematic surface collection across plowed fields; artifact scatters betray buried sites.
- Test pits and sampling — because no one can dig everything, statistics decides where: random sampling avoids bias, stratified sampling covers every terrain zone, and judgment targets the anomalies.
The site is then gridded — a permanent coordinate system so that every find has an address: square, layer, and three-dimensional position. Provenience is the find's meaning; an object without context is merely an objet d'art.
Taking It Apart in Order
Modern practice excavates in single contexts, peeling depositional events in reverse order — youngest first, each context drawn, photographed, levelled, and sheet-recorded before removal, then its finds bagged under the context number. The alternative of digging in arbitrary fixed spits is faster and blinder: a 20-centimeter spit can chew through three floors and mix their contents permanently.
Why not dig it all? Even with perfect recording, excavation spends the site. Best practice leaves part of every site untouched — future archaeologists will bring questions and instruments we cannot imagine, and the unexcavated half is our gift to them. (Radiocarbon, DNA, and isotope analysis all arrived after most of the world's great sites were already dug.)