The Gold Standard of Anthropology | Ethnographic Comparison, Commensurability and Category Imposition

1. Comparison begins with a common question, not a common label

Back to contents

Two ethnographies may both use the English word “household” while describing different social arrangements. One may follow co-residence, another pooled food, another land rights, and another the people named as dependants. A shared label is not yet a shared unit.

Write the comparison as a question about a relation or process: who shares resources, who can claim care, who moves together, or who is counted for an obligation? That common question makes cases comparable without pretending their local categories are identical.

2. Commensurability is designed and bounded

Back to contents

Cases are commensurable when a stated feature can be examined across them with enough equivalence to support the intended inference. The equivalence can be functional, relational, temporal or measurement-based. It need not be complete.

A comparison may support “both arrangements organise daily food sharing” while not supporting “both are the same kind of family.” State the level at which the bridge works and the dimensions it leaves local. Partial commensurability is often the honest result.

3. Keep emic and analytic categories in view together

Back to contents

Emic categories organise experience from within a setting. Analytic categories are constructed to answer a comparative question. The two can inform one another, but neither should disappear into the other.

Build a category map with the local term, literal translation, speakers who use it, contrast terms, ordinary examples, boundary cases and the proposed analytic code. A single English gloss is too thin to show whether two local terms perform comparable work.

4. Fix time, place and case before coding

Back to contents

Cultures and institutions change. HRAF guidance stresses that cross-cultural comparisons need particular time and place foci. A practice described before wage labour, war or migration cannot be merged casually with a later description of the same named community.

Define the unit: person, domestic group, neighbourhood, event or institution. Record observation dates, source type and the event or season represented. When using earlier codes, match the same time and place focus or treat the mismatch as measurement error.

5. Category imposition produces patterned error

Back to contents

An imposed category can erase a local distinction, exaggerate a small difference or combine independent traits. These are not random slips. They can systematically make one case look deficient, exceptional or more similar to the researcher’s home setting.

Use independent coding of difficult cases, written decision rules and a disagreement log. Inspect whether uncertain codes cluster by language, region, source author or social position. A high agreement percentage does not rescue a category that asks the wrong question.

6. Worked example: comparing household membership

Back to contents

Site A uses a local term for people who regularly eat from one cooking pot. Site B names people with continuing claims on a house and fields, including migrants who send money. If both are coded simply as “household members,” Site B appears to have larger households and more absent members.

The team separates three relations: daily co-provisioning, residential presence and durable property or care claims. A migrant at Site B scores no, no and yes; a visiting worker at Site A may score yes, yes and no. The revised comparison can examine how these relations overlap without declaring either local category the universal household.

7. Compare patterns, distributions and negative cases

Back to contents

A binary code often hides internal variation. Report how frequently a relation occurs, among whom, in which situations and with what exceptions. A claim that a category is “collective” should face cases where members act separately or contest inclusion.

Negative cases help locate the boundary of the comparison. They may reveal that the proposed common mechanism operates only during ritual, only for landholding adults or only in one season. The boundary is evidence, not an inconvenience to delete.

8. Changed-context transfer check

Back to contents

Move the household framework to comparative research on “youth.” The discipline transfers: identify local terms, institutional thresholds, rights, obligations and temporal focus before coding. The content does not transfer automatically.

Legal age, school status, marriage, initiation and economic dependence may cut the life course differently. A universal age band can be useful for one demographic calculation, but it cannot silently stand in for locally recognised adulthood. Report both where the question needs both.

9. Limits and common misconceptions

Back to contents

Comparison does not require a claim that cultures are internally uniform. Nor does respect for local meaning forbid abstraction. The error lies in hiding the transformation from local evidence to analytic code.

Do not treat translation as word replacement, use national borders as automatic cultural units, or infer absence because an ethnographer did not mention a trait. Do not call two measures equivalent merely because their response options have the same number of boxes.

10. Audit the comparison chain

Back to contents

Trace each headline result back through analytic code, coding rule, local passage and source context. Recode a sample blind to the expected pattern. Review cases with missing evidence separately from cases coded as absence.

Run sensitivity checks with plausible alternative category boundaries. If the conclusion changes when one disputed relation is recoded, report that dependence. Preserve the codebook, source excerpts permitted by ethics, coder decisions and the version of every derived variable.

11. Frequently asked questions

Back to contents

Must every local term have an exact equivalent?

No. Exact lexical equivalence is rare and often unnecessary. The required equivalence is the one needed for the stated question, and its limits should be visible.

Can researchers use standard categories?

Yes, when the standard category is justified, operationalised and accompanied by checks of how it maps onto local meanings and cases.

Does more coding reliability guarantee validity?

No. Coders can agree perfectly on a poorly designed rule. Reliability checks consistency; validity asks whether the rule represents the intended construct.

12. The gold standard makes the bridge inspectable

Back to contents

Strong ethnographic comparison does not flatten difference. It identifies a common analytic surface, retains the local structures beneath it and tests where the mapping fails.

Before accepting a comparison, ask what the case is, which relation is shared, how local terms became codes, and which negative cases limit the claim. When those steps are visible, comparison can reveal both patterned similarity and consequential difference.