The Archaeological Record and Research Questions · Lesson 5

Evidence register and bounded claims

7 min read

Lesson objectives

  • Explain Evidence register and bounded claims with the evidence limits stated.
  • Produce a reviewable decision from fictional inputs.
Reading tools

After classifying finds, drawing relative sequence, and checking the sample, a further problem remains: how can a reader audit the conclusion? A declaration that “this was an industrial site” followed by several photographs is not a chain of reasoning. The reader needs the source of each observation, its integrity, the inference it is used for, and an alternative explanation. Build an evidence register with separate fields for identifier, context, raw description, recording method, context relationship, candidate claim, confidence and reason, and next test. Separation prevents an interpretation from being entered as an observation and later counted as independent support for itself. Recording guidance uses unique contexts and documented relations; research frameworks connect gaps, questions and methods. The register used here is a teaching device linking those principles, not a mandatory official form. [Historic England, Archaeological Recording Manual, 2018; Historic England, Research Frameworks, checked 2026-10-23]

Confidence is not decorative. State its basis: “high confidence that this cut is later, because its edge is visible in the section and photograph,” or “low confidence linking the sherd to this phase, because it came from mixed fill.” Evidence can be securely located yet weak for assigning a room's function. Distinguish uncertainty that another record may reduce, such as a missing image that might be recovered, from a limit of the available material, such as an old boxed object with no documented findspot. General methodological references support principles of interpretation; they do not prove the invented case's observations. Only a case record can support a statement that something was seen there. If that record is missing, the claim must remain hypothetical.

For each claim, ask whether a rival explanation fits the same observations. Distorted sherds and a burn patch may fit production, but they may also be waste imported beside a domestic hearth. Seek evidence that differentiates them: production debris in an intact relation to a fixed firing feature, or only malformed pieces in mixed fills? Then calibrate the sentence to the evidence. “The record raises a production possibility that requires a context check” is more defensible than “we proved a factory.” New images or analysis may change the assessment. Record the date, reason and version of a changed decision rather than deleting its predecessor, so another reader can understand how the interpretation evolved.

A review meeting may reveal ten gaps when there is time to resolve only a few. Rank them by effect on the decision and the possibility of verification. A sample-unit error may overturn the sequence, whereas a missing photographer credit may affect attribution and rights. Both matter, but priorities depend on the question. For each gap, list affected claims, the source that could resolve it and the person able to retrieve that source. Distinguish obtaining an existing document from commissioning new research. Opening an old field book may restore a missing link quickly; identifying a material may require prolonged specialist study. A practical follow-up recommendation states its expected output and how that output would change the report. This is a training exercise in evidence quality; actual responsibilities remain institutional.

Lesson summary

After classifying finds, drawing relative sequence, and checking the sample, a further problem remains: how can a reader audit the conclusion? A declaration that “this was an industrial site” followed by several photographs is not a chain of reasoning.