Archaeology · field methods
Excavation is controlled destruction. You only get to dig a site once, so the record you make is the site.
An artifact on its own tells you almost nothing. An artifact in context — this layer, this depth, next to that hearth, under that floor — tells you when, who, and how. Everything in field method exists to protect that context, because once you dig it out, it is gone for good.
Excavation is destruction. You get one pass through a deposit, and everything you fail to record is gone. So the work that matters most happens before a trowel touches soil — and most of it is not glamorous.
The standard CRM sequence in the United States. Each phase answers one question and only proceeds if the answer requires it.
| Phase | Question | Typical methods |
|---|---|---|
| Phase I | Is anything here? | Background review, pedestrian survey, systematic shovel tests on a grid, sometimes geophysics |
| Phase II | Does it matter? — testing for eligibility | Larger test units, enough excavation to assess boundaries, depth, integrity and significance |
| Phase III | Mitigation — recovering what will be lost | Large-block data recovery, full analysis, reporting and curation |
Phase III means the site is going to be destroyed and excavation is the consolation prize. Avoidance is always the better outcome — redesigning a project around a site preserves it, and preservation in place beats even the best data recovery. Most projects stop at Phase I, which is the system working as intended.
Law & process
National Register criteria
A property may be eligible if it:
A — is associated with important events
B — is associated with important people
C — embodies distinctive design or construction
D — has yielded, or may yield, information important in history or prehistory
Archaeological sites are usually evaluated under Criterion D. Eligibility also requires integrity, assessed across seven aspects: location, design, setting, materials, workmanship, feeling and association. A site that has been thoroughly ploughed, looted or bulldozed may be genuinely important and still fail on integrity, because the information Criterion D depends on is no longer recoverable.
Non-invasive methods do not replace excavation, but they tell you where to put a unit — and sometimes they answer the question without breaking ground at all, which is always the better result.
| Method | What it senses | Good for | Limits |
|---|---|---|---|
| Ground-penetrating radar | Reflections from contrasts in buried material | Walls, voids, graves, pit features; gives depth | Struggles in wet clay and high-conductivity soils |
| Magnetometry | Local distortions of the magnetic field | Hearths, kilns, burnt features, iron, filled pits — fast over large areas | Wrecked by modern iron, fences and power lines; no depth |
| Electrical resistivity | How readily current passes through soil | Stone walls (high resistance) and ditches (low) | Slow; strongly affected by recent rainfall |
| LiDAR | Laser returns from the ground surface | Microtopography under vegetation — mounds, causeways, field systems | Surface relief only; sees nothing buried flat |
| Satellite & historic aerials | Crop marks, soil marks, vanished landscapes | Wide-area reconnaissance and change over time | Resolution and seasonality; needs ground truthing |
The consistent lesson is that these methods are complementary, not interchangeable. Magnetometry and GPR over the same field routinely find different features, because they are sensing different physics. And every anomaly is a hypothesis until something confirms it — geophysics tells you where the contrast is, not what caused it.
A 1 × 1 m test unit at the Ridgefield Site. Take it down level by level and make the right call at each one — you cannot put a layer back.
Level 1 · 0–10 cm
Loading…
Excavation destroys its own evidence. The trench is gone the moment you finish it, and what survives is the record you made while it existed. Note-taking is genuinely personal — every archaeologist’s book reads differently — but the structure underneath is not optional, and nobody is born knowing it.
Write this list inside the front cover of your field book and fill it out at the top of every working day. It takes ninety seconds, and it is what makes the rest of the page usable by someone who was not there.
Both belong in the book. What must never happen is a reader being unable to tell which is which — because your interpretations will sometimes be wrong, and when they are, the observations underneath still have to survive.
Reproducible. Someone standing beside you would write something close to the same thing, and a reader can check it later against the section drawing, the photographs and the bags.
10YR 3/2 very dark greyish brown silty loam, friable, frequent charcoal flecks <5 mm. Lower boundary sharp, level, 46 cm BD. Screened 1/4 in.A reading of the evidence. Legitimate, necessary, and to be written down — but marked as yours. Indent it, bracket it, prefix it; the convention matters less than the consistency.
INT: charcoal concentrated rather than dispersed, boundary flat — possibly in-situ burning rather than redeposited rake-out. Check against Feature 7 fill.This looks like data and is not. There is no defined scale, so nobody knows what 7 means or what it is 7 of. There is no reference standard, so the value cannot be reproduced by another recorder, or by the same recorder next week. And it records a judgment in place of the observation, so the evidence behind it is gone. The one useful thing on the page — that the deposit is sand — is the only word with no number attached.
The lesson is not that qualitative or ranked recording is wrong. Archaeology and soil science run on ordinal scales and they work, because they are anchored: Munsell fixes colour against a physical chart, USDA texture classes are defined by measurable grain-size proportions, and abundance scales such as rare / occasional / frequent / abundant come with stated definitions. The test is simply — could someone else reproduce this reading, or check it? If not, it is not a record. It is an opinion with a number stapled to it.
Sort real field-book lines. Some are solid records, some are readings of the evidence, and some cannot be used for anything.
Sort it
Mechanics that save books
Pencil, or archival pen — ballpoint fails in rain and gel ink runs. Waterproof paper if the budget allows.
Never erase, never tear out a page. Strike through with one line, write the correction, initial and date it. The struck-out version is part of the record, and sometimes the interesting part.
Write in the unit, not from memory at dinner. Detail decays within the hour, and what goes first is whatever you did not yet know mattered.
Sketch badly rather than not at all. A rough plan with a north arrow and a scale beats a perfect memory. Add both before you add anything else.
Photograph or scan the book each night. It is the one irreplaceable object on site, and it lives in a bag that goes in a truck.
Each entry below is reasonable as far as it goes, and each omits one thing that makes it unusable later. Find it.
Spot the gap
The test to apply to any line you write
Could a stranger, in twenty years, holding your book and nothing else, reproduce your reasoning and check your evidence?
If yes, the line is doing its job. If it depends on something only you saw, only you remember, or only you can interpret, then whatever that missing thing is needs writing down next to it. That is the whole of field recording — everything above is just the standard places people find themselves failing that test.
Three words do most of the work in archaeology: artifact, ecofact, feature. Add context and you have the whole discipline in four terms.
Classify it
Loading…
First work out the order, then pin it to calendar years. Relative sequence comes from the ground; absolute dates come from the lab — and each one dates a specific thing, not the event you care about.
Contexts from a section at the fictional Quebrada Verde site. Click them earliest first — cuts, fills and layers all take their place in one sequence.
Which context is earliest?
The question
Loading…
Some objects date a deposit almost on sight, because their manufacture changed on a known schedule. Nails are the workhorse on historic sites; temper does similar work on prehistoric pottery. Read the attributes and name it.
Identify it
Nails, in three steps
Hand-wrought — to c. 1800. Forged one at a time: shaft square and irregular, tapering on all four sides, head hammered on by hand and never quite even. Soft iron, so they survive bent rather than snapped.
Machine-cut — c. 1790s–1890s. Sheared from rolled plate: two parallel faces of uniform thickness with the taper only on the other two, and burred edges from the shear. Early ones still had heads applied by hand and are irregular; after about 1830 the heads are machine-made and regular.
Wire — common from the 1890s. Cut from round drawn wire, with a stamped round head. Once you see round shafts, you are in the twentieth century or very close to it.
Square shaft means pre-1890s; four-sided taper and an uneven hand-made head push it earlier still. Nails are abundant, and a handful gives a far better date than one of anything else.
Excavation produces numbers, and the numbers have to be made to say something. Three of the workhorse calculations — dating a deposit from its pottery, calibrating a radiocarbon age, and reading a season out of bone and seed.
Stanley South’s formula, and still the standard first pass at dating a historic-period deposit. Every ceramic type was manufactured over a known span; take the midpoint of that span, weight it by how many sherds of that type you found, and average.
| Ceramic type | Manufacture span | Midpoint | Sherds | Midpoint × count |
|---|
Interpretation
Two dates, two jobs
The MCD estimates the middle of the period the deposit accumulated — it is an average, so it is pulled by everything in the sample.
The TPQ — terminus post quem, “limit after which” — is set by the youngest item present: the deposit cannot have formed before that type existed. One is an average, the other is a hard floor, and they answer different questions. A single intrusive modern sherd wrecks the TPQ while barely moving the MCD.
A radiocarbon measurement is not a calendar date. Atmospheric 14C has varied over time, so the measurement has to be calibrated against a curve built from tree rings. Where the curve is steep, a measurement pins down a narrow calendar range. Where it flattens or doubles back, the same measurement sprawls across centuries — and nothing about a better lab or a smaller error bar fixes that.
The curve above is schematic — drawn to show the shape of the real feature, not to calibrate anything for publication. That feature is genuine: the Hallstatt plateau, roughly 800–400 cal BC, where atmospheric 14C rose enough to hold measured ages nearly flat for four centuries. It is the reason Iron Age and Scythian material so often dates to “somewhere in the first half of the first millennium BC” and no closer. Real work uses IntCal and reports probability distributions, not a single range.
Plant and animal remains carry the season they were acquired in. Assemble enough indicators and you can say when a site was occupied — and whether people stayed all year or came back at the same time each year.
The kit, the paperwork, and the law.
Everything between deposition and your trowel — and why the ground is never a simple layer cake.
An artifact removed from the ground without record loses almost everything that made it evidence. The object survives; the context — its layer, its associations, its date, its relationship to everything around it — is gone permanently, and no amount of later study recovers it.
That is why professional practice treats stewardship as the first obligation, why site locations are kept out of public records, and why buying antiquities funds the destruction of exactly the information archaeology exists to recover.
It is also why descendant communities are consulted rather than informed. The past being excavated belongs to people who are still here.