A heading says that a society possessed roads. Another says that it had schools. A third calls it commercially advanced. The profile looks complete until someone asks three ordinary questions: which roads, what counts as a school, and what evidence supports advanced?

A civilisation profile should not be a collection of impressive labels. It should be a set of claims that a reader can trace, question and revise.

The first article in this series, How to Categorise Civilisation?, proposed a multidimensional approach. This article turns that idea into a practical research exercise. It explains how to define the object, select evidence, handle missing information, compare measures and write a conclusion that remains within the record.

The worked societies and records below are fictional. They are teaching constructions, not newly discovered historical evidence. The methodological references come from Seshat’s description of its research process and the European Commission’s work on indicators. The worksheet is an original educational application, not an official Seshat coding instrument or a validated universal civilisation index.

Begin with a question small enough to answer

What was this civilisation like is a reasonable opening question, but a difficult research specification. It contains too many possible objects, periods and standards of judgement. A more useful starting point might be: how did two urban communities organise access to water during a defined period?

That question does not reduce civilisation to water. It establishes a manageable entry point into infrastructure, institutions, labour, knowledge, geography and household experience. Other dimensions can be added when they help answer the question rather than because a template contains an empty box.

A profile should state its reader job. Is it introducing a historical society? Comparing one capability? Explaining a transition? Preparing questions for further research? Different jobs require different evidence. A broad introduction may responsibly remain descriptive where a causal claim would need much more support.

This distinction protects the reader from a common escalation. A descriptive fact becomes a category, the category becomes a score, and the score becomes an explanation of history. Each step adds a claim. None is automatically licensed by the previous one.

Give the object a boundary before giving it a label

Write down the place, period and unit of analysis. A city, a polity, a trading network and a cultural tradition are not interchangeable units. A polity means a political entity; a civilisation label can cover several such entities across a much longer period.

In our fictional exercise, River Plain refers to one urban community during Years 90–110 of an imaginary chronology. Harbour refers to another urban community during the same interval. The names are placeholders. Neither stands for a real people, ethnicity or modern state.

This boundary prevents evidence about a surrounding kingdom from silently becoming a claim about the city. It also prevents a record written much later from being treated as a direct observation within the chosen interval. Later records can still be relevant, but their relationship to the period needs explanation.

Seshat’s methods distinguish spatial sampling, political entities, time series and the coding of particular variables. Its process also records sources, uncertainty and scholarly disagreement. The relevant lesson here is that historical comparison requires explicit units and traceable decisions, not simply familiar civilisational names. [1]

Separate the object, the evidence and the interpretation

Suppose a stone inscription says that a landing stage was repaired. The object is the inscription. The evidence includes its wording, location, date and relationship to other material. The interpretation might be that somebody claimed responsibility for maintaining a transport facility.

That is not yet proof that the facility operated reliably for the whole population. The inscription may commemorate an exceptional event. The work may have been incomplete. The facility may have served only some users. These are possibilities to investigate, not reasons to dismiss the inscription automatically.

A useful profile therefore keeps an evidence statement beside an interpretive statement. The first says what was found or recorded. The second says what the researcher thinks it supports. A third field can state what it does not establish.

This makes the argument inspectable. Readers can agree that the inscription exists while disagreeing about how far its implications extend. Without that separation, disagreement about interpretation can be mistaken for disagreement about the evidence itself.

A small fictional dossier

For River Plain, imagine three constructed records. A ledger dated within the study interval lists deliveries to three grain depots. A site survey identifies ten public water points. A population estimate places the urban population at 40,000 for the chosen calculation. The exercise treats that estimate as fixed initially, then examines uncertainty later.

For Harbour, imagine a contemporary inscription commemorating repairs to a landing stage, a site survey identifying six public water points and a population estimate of 20,000. A later chronicle, written forty years after the study interval, describes schools but provides no clear evidence of how many existed during the selected years.

Each record has a different job. The grain ledger supports a claim about recorded movement through named depots. The landing-stage inscription supports a claim about a contemporary commemoration of maintenance. The water-point counts support an inventory measure under the survey’s definitions. The later chronicle introduces a question about education; it does not settle the answer.

The dossier is deliberately incomplete. The purpose is to learn how to work with incomplete evidence without making the profile falsely complete. A blank area in the record should create a research question, not an invitation to fill the space with a plausible story.

Write a definition that another researcher could apply

A category such as infrastructure is too broad to code consistently without further definition. Does a road include an informal path? Does a school mean a building, a teaching arrangement, an institution named in a text or any place where learning occurred?

Seshat’s published road variable illustrates this issue. It distinguishes deliberately constructed routes connecting settlements or other sites from internal streets and paths formed by repeated use. The exact boundary matters because different definitions would produce different counts. [2]

For our exercise, define a public water point as a location counted by the fictional survey as available for public collection during the specified period. Immediately record the limits: the definition does not prove continuous operation, equal access, safe quality or sufficient quantity.

The definition may later need revision. If it changes, earlier coding should be reviewed rather than silently left under the old meaning. Otherwise two entries can share a column heading while actually measuring different things.

Do not make absence of evidence do more work than it can

The fictional Harbour chronicle mentions schools after the selected interval. We do not yet have adequate evidence to describe schooling during Years 90–110. The appropriate entry is unknown for that question, with a note about the later source.

Writing no schools would make a stronger claim. It would suggest evidence for absence rather than a gap in our knowledge. Conversely, writing established education system would convert a later account into a confident description of an earlier period.

Seshat explicitly distinguishes present, absent and unknown, with separate inferred categories when the evidence is indirect. It also records ranges and disagreements where relevant. [1] The purpose is not to eliminate judgement but to make the type of judgement visible.

Evidence of absence is possible in some settings. A sufficiently complete record created for a relevant purpose might support it. But the reasoning must explain why the expected evidence would probably have appeared if the feature existed. An unsuccessful search alone does not establish that condition.

Use uncertainty labels that describe the evidence, not your confidence as a writer

A fluent paragraph can make a weak inference sound strong. To prevent that, use plain labels such as directly documented, indirectly inferred, disputed, unknown and not applicable. Define them for the project. They are not interchangeable ways to express hesitation.

Directly documented does not mean beyond question. A document may directly state something that was exaggerated or mistaken. The label describes the relationship between the claim and the source, not a guarantee of historical truth. A further assessment of source quality is still needed.

Disputed means there are identifiable competing interpretations or findings. Unknown means the available record does not currently support a decision. Not applicable means the variable does not meaningfully fit the defined case. Treating all three as zero would erase distinctions that matter to the analysis.

In the fictional dossier, the grain movement is documented by the invented ledger, while the reliability of the whole food-distribution system remains unknown. Those statements can coexist. A civilisation profile becomes more informative when it preserves such distinctions rather than insisting that every dimension receive one unqualified adjective.

Ask whether the source could know what you are asking it to prove

A tax record may be useful for particular administrative obligations. It may be a poor source for the experiences of people who were not recorded. A monument may reveal how an authority represented itself. It may not reveal whether ordinary households experienced the same authority as fair.

The question is not whether a source is simply good or bad. It is whether it is suitable for the claim being made. Who produced it? For what purpose? What could the producer observe? What incentives might shape the account? Which people or activities were outside its field of view?

Apply those questions to the fictional grain ledger. It can show that specified deliveries were recorded. It cannot establish that every household received sufficient food unless additional evidence connects the ledger to household access. A record of organised distribution is not automatically a record of universal provision.

This step is especially important when the source is impressive. Age, official appearance or technical detail can make an object feel authoritative. Authority for one question does not automatically transfer to every adjacent question.

Corroboration is not the same as repetition

Imagine that three modern summaries repeat the same account of Harbour’s landing stage. If all three depend on one inscription, they do not provide three independent observations of the repair. They provide three pathways back to one source.

This does not make the summaries useless. They may offer different expertise or interpretations. But a profile should distinguish independent evidence from repeated transmission. Otherwise the number of references can create an illusion of confirmation.

A stronger dossier might combine the inscription with independently dated physical work, records of expenditure or other contemporary observations. Whether those materials are genuinely independent and consistent would still need investigation. More sources can introduce more complexity rather than simply adding certainty.

For students, a practical habit is to write one sentence about the origin of each major claim. When several references share the same origin, note that relationship. The aim is not a bibliography of maximum length, but an account of what evidence actually bears the weight.

Counts need denominators—and denominators need boundaries

River Plain has ten public water points. Harbour has six. On the raw count, River Plain has more. With the fictional populations of 40,000 and 20,000, River Plain has 2.5 counted points per 10,000 residents, while Harbour has three.

The calculation is ten divided by 40,000, multiplied by 10,000, compared with six divided by 20,000, multiplied by 10,000. The ranking changes because the question changes from total count to count relative to population.

Neither figure establishes that Harbour had better water access. A point may have different capacity, reliability, operating hours or location. People may travel different distances or face different restrictions. The per-capita count is a defined indicator, not the whole capability.

The European Commission’s indicator guidance emphasises relevance, measurability, coverage and analytical soundness when selecting measures. A readily available number is not automatically a suitable indicator of the phenomenon the researcher wants to describe. [3]

A population estimate should not become invisible certainty

Now replace the fixed River Plain population with a fictional range of 35,000 to 45,000. The same ten water points give approximately 2.86 to 2.22 points per 10,000 residents. For Harbour, a fictional population range of 18,000 to 24,000 gives approximately 3.33 to 2.50 points per 10,000.

The ranges overlap. Our narrow claim that Harbour certainly had the higher per-capita count is no longer supported across all combinations of those estimates. The appropriate conclusion should reflect that uncertainty.

This is an elementary sensitivity check: vary an uncertain input and see whether the conclusion changes. It does not assign probabilities to the endpoints or prove that every value in the interval is equally plausible. It simply shows which parts of the result depend on the chosen assumptions.

The Commission’s discussion of composite indicators similarly includes missing data, bringing unlike measures into comparable units, weighting and checking robustness through uncertainty and sensitivity analysis. [4] Our example applies only the basic sensitivity idea to a transparent ratio.

Do not treat assets as proof of dependable capability

A profile can record that an asset existed without claiming that it worked continuously. Ten water points are not ten reliable water services. A road on a map is not proof that goods moved along it safely throughout the year. A named office is not proof that it performed its stated role.

To move from asset to capability, look for evidence of use, upkeep, staffing, resources, accessibility and performance. The relevant evidence will differ by domain and historical setting. The important thing is to recognise that an additional claim is being made.

For River Plain, the fictional survey supports an asset count. It does not establish maintenance capacity. For Harbour, the inscription offers a claim about a repair, but not a complete service history. Neither society should receive a confident reliability label merely because the profile contains an infrastructure row.

This connects to capability across generations. Civilisation is not only what was built. A profile that aims to describe reliable action needs evidence about what could be operated, maintained and reproduced.

A profile is not yet a causal explanation

Suppose Harbour appears more connected to maritime trade and also has a higher count of a particular institution. That association does not establish that trade caused the institution. Both may have been influenced by other conditions, or the direction of influence may run both ways.

A descriptive profile says what the evidence supports about each dimension. A causal explanation asks why one condition produced another, compared with a relevant alternative. That is a different research task.

Students can preserve this boundary by changing their language. The records are consistent with is weaker than the records prove. These features occurred together is different from one caused the other. A possible explanation is not the same as an established mechanism.

The Atlas already connects to broader methods in How Causal Inference Works. The profile’s job is to prepare a defensible description and reveal good questions, not to claim that a few columns settle the causes of civilisational development.

Why a single score can conceal the most important finding

Imagine combining infrastructure counts, written records, political scale and cultural production into one score. Before calculation, the researcher has already made choices about which dimensions matter, how they are measured and how much weight each receives.

A composite can be useful for a clearly defined purpose. It should not be mistaken for an objective measure of a society’s total worth. Nor should one good result automatically compensate for an unrelated serious harm simply because the arithmetic adds the dimensions together.

In the fictional exercise, the most useful finding may be that evidence is strong for some infrastructure assets and weak for household access. Averaging those dimensions could hide the research gap. A profile preserves it.

Missing values create another problem. If unknown schooling is treated as no schooling, the score embeds an unsupported historical claim. If it is silently omitted, different societies may be scored on different sets of evidence. Either choice must be explained rather than buried inside the formula.

Write the conclusion at the same resolution as the evidence

A weak conclusion would say that River Plain was a superior civilisation because it had more infrastructure. It combines an asset count, an undefined idea of superiority and a judgement about an entire society.

A stronger conclusion is narrower: within the fictional Years 90–110 study window, the constructed surveys record more public water points in River Plain, but the fixed-population calculation gives Harbour a higher count per resident. Allowing the specified population ranges makes that comparison uncertain. Neither inventory establishes service reliability or equitable access.

That conclusion may sound less dramatic. It gives the reader much more usable information. It identifies what was compared, the evidence boundary, the effect of the denominator and the questions not answered.

A second paragraph can identify next research steps: seek evidence of operating capacity, maintenance, access and the distribution of households. Those are not ceremonial caveats. They explain what information would most improve the profile.

A reusable profile record

For each dimension, record the claim in plain language, the place and period, the variable definition and the unit of measurement. Add the source and the specific observation it supplies. Then state the interpretation, uncertainty and important limitations.

A worked entry for River Plain might read: ten public water points counted in the fictional urban survey for Years 90–110; population initially assumed at 40,000; calculated density 2.5 points per 10,000 residents; operation, quality and household access not established. The fictional estimate range changes the density to approximately 2.22–2.86.

Another entry might read: grain deliveries to three depots recorded in a contemporary fictional ledger; supports organised recording of specified transfers; does not establish the total food supply or its distribution to all households. No population-wide adequacy claim is made.

The record should include when it was reviewed and what could change the judgement. A later discovery should not require rewriting the whole civilisation from memory. It should be possible to identify which claim depends on the affected evidence.

Let another reader try to break the profile

Give the profile to someone who did not prepare it. Ask them to reproduce one calculation, apply one definition to a new example and locate the evidence supporting one major sentence. This is a classroom review exercise, not a substitute for scholarly peer review.

If the reader cannot reproduce the result, find out why. The source may be unclear, the denominator may be missing or the claim may depend on an assumption that was never written down. These are useful discoveries.

Ask the reader to propose an alternative interpretation that fits the same evidence. The aim is not to treat every interpretation as equally strong. It is to discover whether the preferred conclusion has been tested against a plausible competitor.

Finally, ask which statement sounds more certain than the record allows. That question often improves a profile more than adding another page of description. It keeps the evidence and the language at the same level.

The profile is a version, not the final civilisation

A civilisation is larger than the surviving record, and the record is larger than any one profile. New evidence, improved dating, revised definitions or a better question may change the result.

That does not make careful classification pointless. It makes traceability valuable. A profile that explains its boundaries and reasoning can be revised without pretending that the earlier version never existed. A profile built from unqualified labels gives future readers much less to work with.

Our fictional societies remain incomplete at the end of the exercise. We know more clearly what the constructed evidence supports, where the comparisons are fragile and what further information would matter. That is a successful result for this research job.

A good civilisation profile does not claim to contain an entire human world. It shows exactly which part of that world the evidence currently lets us describe—and leaves a clear path for the next reader to understand more.

Sources and scope

[1] Seshat: Global History Databank, How We Work. Used for its description of explicit variables, time and spatial units, scholarly provenance, uncertainty, inferred coding and recorded disagreement. No Seshat dataset has been copied into the fictional examples.

[2] Seshat, Road variable definition. Used as a concrete example of how an operational definition distinguishes constructed connecting roads from internal streets and paths created through repeated use.

[3] European Commission Knowledge4Policy, Step 2: Data selection. Used for the importance of relevance, measurability, coverage and analytical soundness in selecting indicators.

[4] European Commission Knowledge4Policy, Tools for Composite Indicators Building. Used for the need to examine missing values, normalisation, weighting and robustness when combining indicators.

River Plain, Harbour, their records, the chronology and all quantities are invented. The worked arithmetic demonstrates a method; it is not evidence about any actual historical society. Categories are used to describe systems, not to rank the dignity or worth of peoples.

Continue through the Civilisation Atlas

Return to How to Categorise Civilisation?, connect the profile to measuring progress and decline, or browse the Civilisation Atlas. For the broader comparison method, continue to How Comparative Systems Research Works.