2 Uncertainty and Inference

Why conclusions are uncertain

An inference is a conclusion drawn from evidence. Evidence rarely determines only one possible explanation: observations may be incomplete, measurements imperfect, and the processes that produced the evidence only partly understood. Uncertainty is therefore not automatically a sign of poor research. It is a feature of reasoning from limited evidence and should be described rather than concealed.

Several sources of uncertainty can affect the same conclusion:

• Measurement: Instruments, dating methods, and observers have limits or may introduce error.

• Sampling: The evidence collected may not represent the full population, place, or period being studied.

• Interpretation and models: Different explanations or analytical choices can fit the same observations.

• Variation and chance: Real processes vary, so repeated observations may differ.

• Missing or unevenly preserved evidence: What survives may not reflect what once existed. In archaeology, some materials decay more readily than others. A small number of surviving plant remains cannot, by itself, show how common each plant was in the past.

• Research and documentation bias: Some places, people, or materials may have been studied or recorded more often than others.

These sources can combine. A claim about past activity may depend both on whether an object was correctly identified and on whether the excavated area represents the site as a whole. Researchers should identify the main uncertainties bearing on a conclusion and explain how they limit it.

Assessing confidence and probability

Confidence describes how well supported a conclusion is. It is an overall judgment about the strength and consistency of evidence and its interpretations. Probability, by contrast, describes how likely a clearly specified outcome is. These concepts are related but not interchangeable: “high confidence” does not automatically mean a particular percentage chance.

To assess a claim, consider its evidence and the limits on what that evidence establishes:

1. Examine the evidence. Consider its relevance, quality, amount, and how it was collected.

2. Check independence. Several measurements that rely on the same assumptions may provide less support than genuinely different methods.

3. Assess consistency. Note both agreement and important conflicting evidence.

4. Consider what may be missing or biased. This includes preservation, sampling, recording, and access to evidence.

5. Compare explanations. Ask how strongly the evidence distinguishes the proposed explanation from alternatives. A plausible interpretation is not necessarily the best-supported one.

Scientific assessments commonly evaluate the type, amount, quality, and consistency of evidence, as well as the degree of agreement among lines of evidence. Multiple independent, consistent lines of high-quality evidence generally support stronger confidence than a single, narrow source.

Some fields use calibrated labels such as low, medium, or high confidence. Such labels are useful only when their meaning is defined and applied consistently. The IPCC distinguishes qualitative confidence in a finding from calibrated terms for the probability of a specified outcome. Its probability ranges belong to that assessment framework; they are not universal meanings for words such as likely or very likely.

Using probability carefully

Do not turn a qualitative judgment into a precise percentage unless the evidence and method justify that precision. If a probability is estimated, state what event it concerns, how the estimate was made, and any relevant range or assumptions. “We do not know” is not equivalent to saying an outcome is equally likely either way: uncertainty may reflect a lack of evidence rather than balanced support for competing possibilities.

Communicating conclusions responsibly

A responsible explanation makes clear what is observed, what is inferred, and how secure the inference is. It also identifies the limits of the evidence and reasonable alternative explanations.

For example, charred seeds from an excavated area show that a plant was present at the site. Because plant remains preserve unevenly and the sampled area is limited, those finds do not establish how common the plant was across the settlement. Saying “the plant was rare” would overstate what the sample alone establishes.

When communicating a conclusion:

• State it at the strength the evidence supports—not more strongly, and not needlessly more weakly.

• Name important limitations and competing explanations rather than hiding them in technical detail.

• Use probability terms only when their meaning is clear; give numerical ranges where appropriate and justified.

• Distinguish evidence about a claim from disagreement about its meaning. Do not present unequal levels of support as equal merely to sound balanced.

• Explain what new evidence or analysis could change the conclusion. Knowledge can be revised as evidence and methods improve.

Ethical responsibilities

How evidence is collected, interpreted, preserved, and shared affects both knowledge and people. Researchers should document methods and decisions, report results accurately, preserve records and collections for future work, and acknowledge relevant contributions. Archaeological reporting also requires attention to descendant and Indigenous communities’ knowledge and interests. Responsible reporting and collaboration matter, and sharing sensitive information, such as site locations, can have unintended consequences.

Ethical care is especially important when evidence concerns human remains, sacred places, or culturally sensitive knowledge. Researchers should follow applicable law and relevant professional guidance, consult affected communities, and avoid presenting uncertain interpretations as settled facts about people or communities. Acknowledging uncertainty does not remove the responsibility to make careful judgments; it helps others understand the basis and limits of those judgments.