3 Preservation Bias

What preservation bias means

Preservation bias is the systematic distortion of evidence caused by the uneven survival, selection, accessibility, or description of materials. The evidence available to researchers is not a complete or neutral sample of what once existed; it has passed through successive filters: what was created, what survived, what was collected, what was retained, what can be accessed, and what can be found through a catalog or search system.

Bias is not simply a flaw in evidence. It is a feature of how evidence became available, and its importance depends on the research question. A fossil assemblage that preserves large, hard-bodied organisms may be useful for studying those organisms but poor evidence for reconstructing an entire ancient ecosystem. Archaeological preservation, recovery strategies, and analytical choices also affect what can be inferred from recovered material.

Survival and selection

Materials differ in their chances of survival. Climate, burial conditions, fire, decay, reuse, and later construction can preserve some objects or records while destroying others. Durable stone or fired ceramic may survive where textiles, food remains, or paper do not. Survival can vary within a single site or collection, so the surviving sample may overrepresent particular places, periods, or kinds of activity.

People decide what to collect, excavate, retain, or discard. Collection and selection reflect research questions, cost, available technology, institutional priorities, and judgments about what is valuable. In archaeology, survey visibility, excavation strategy, screening methods, and sample size influence what is recovered. Museum and archival collections also reflect acquisition choices that may favor particular subjects, creators, communities, or kinds of objects. These choices shape later research even when the original selection was practical rather than deliberate exclusion.

Access and description

Preserved material may still be unavailable because it is physically distant, fragile, uncataloged, restricted by law or agreement, or absent from digitized collections. Digital access is selective too: a searchable online collection may contain only part of the physical holdings. Restrictions may be justified to protect privacy, sacred knowledge, vulnerable sites, or culturally sensitive materials. Researchers should not mistake inaccessible evidence for nonexistent evidence.

Cataloging and description make materials discoverable, but also condense and interpret them. A name, subject heading, date, or description can determine which searches retrieve an item. Outdated or harmful terminology may misrepresent people; inconsistent names may scatter related materials across records; and a collection-level description may conceal important items. Archival description is a user-facing guide to records and their contexts, not a complete representation of every item.

Recognizing bias in evidence

A careful researcher asks not only what evidence shows, but how it became available. Use these checks to assess the limits of an inference:

1. Define the claim and relevant evidence. Specify what kinds of materials would be expected if the claim were true, and which are likely to survive or be collected.

2. Trace the evidence’s history. Check where, when, and how it was created, recovered, acquired, processed, described, and made accessible. Examine provenance, collection policies, excavation or sampling methods, and known gaps.

3. Compare like with like. Consider whether apparent differences between periods or groups could result from differences in preservation, survey effort, collection practices, or cataloging rather than differences in past behavior.

4. Look for missingness patterns. Ask whose records or objects are absent, which materials are underrepresented, and whether gaps cluster by community, place, time, format, or institutional policy. A gap is a clue to investigate, not by itself proof of absence or deliberate suppression.

5. Use independent evidence. Compare collections and repositories, and, where appropriate, combine material, documentary, oral, environmental, or community-held evidence. Different sources can have different biases; agreement is informative but does not automatically remove uncertainty.

6. Record limits and uncertainty. State what was searched, what was unavailable, and how sampling or preservation could affect the conclusion. Do not treat the surviving sample as representative without evidence that it is.

For example, an excavation may yield many pottery fragments but few plant remains. It would be unsafe to conclude that people ate little plant food: plant remains may decay more readily, and recovery may not have included fine screening or flotation. The narrower conclusion is that pottery is abundant in the recovered sample, while plant remains are scarce under the preservation and recovery conditions used.

Ethical responsibilities

Researchers and collecting institutions have responsibilities beyond identifying bias. They should document selection and processing decisions, preserve records and collections for future research, describe uncertainty honestly, and make evidence findable and accessible where appropriate. Stewardship includes consulting and involving descendant and Indigenous communities in decisions about research, collections, access, and culturally sensitive knowledge.

Access is an important goal, but it is not always unrestricted. Privacy, community authority, and protection of vulnerable materials may require limits. When descriptions are incomplete or harmful, institutions can revise them, explain changes, preserve relevant context, and work with affected communities. Ethical correction should improve discovery without concealing the history of earlier collecting or description practices.