The Gold Standard of Linguistic Typology

The gold standard of linguistic typology is a comparison that makes its categories, evidence and limits visible. Linguistic typology studies structural similarities and differences across languages: how clauses organise participants, how words express grammatical information, and which patterns occur together. A useful comparison explains exactly what was compared before claiming that languages share a type.

This root guide starts with a practical question: how can two languages communicate a similar event through different arrangements? It connects linguistics, world languages, morphology, translation and language evidence. Its aim is a method readers can reuse, rather than a list of labels to memorise.

The educational purpose follows The Gold Standard of Education: learning should support sound judgement in unfamiliar situations. All invented languages, sentence collections and numerical datasets below are original teaching exercises. They do not report findings about real communities. The featured study-desk photograph illustrates language study; it is not a language-distribution map.

1. Ask a question small enough to answer

Begin with a feature, a comparison unit and an intended claim. “How do languages work?” is an excellent motivation but too large for one coding decision. “In these descriptions, does a possessive phrase place the possessor before or after the possessed noun?” gives the researcher something identifiable to inspect.

The first principles are straightforward. A comparison needs observations that can be checked. It needs a definition that can be applied consistently. It needs room for evidence that does not fit the initial categories. Finally, its conclusion must match the material actually examined. Skipping any one of those steps makes a neat table less informative than it looks.

Write a question card before collecting examples: target construction, varieties included, period represented, source type and exclusion rules. A project might include ordinary affirmative clauses but exclude quotations, questions and subordinate clauses. Those exclusions shape the answer; they belong beside the result.

A useful classroom opening is to show two invented records describing the same pictured event. Ask what identifies the actor and what identifies the affected participant. Learners may notice position, an ending or a separate marker. Their first observation becomes a hypothesis to test with further records.

The WALS introduction explains the value of comparing structural features across languages. Here the question card is a teaching device for turning that broad purpose into a manageable investigation.

Contents · Next chapter

2. Keep structural type separate from language history

A structural category and a historical relationship answer different questions. A type groups languages according to a chosen property. A family groups them according to historical descent supported by comparative evidence. A shared property alone does not establish common ancestry.

Imagine three fictional varieties. Amber and Birch share a documented ancestor but differ in their current placement of objects. Cedar resembles Amber in that placement, yet the exercise supplies no historical link between them. The correct table can record both facts: Amber and Birch belong together historically, while Amber and Cedar share the selected structural value.

Now add geography. Nearby communities might share features through contact, while relatives separated by distance may develop differently. Similarity therefore creates questions about inheritance, interaction and other explanations. It does not choose among them by itself. A coloured cluster on a map is an invitation to investigate.

The Glottolog explanation of languoids and classification provides a useful reference for identifying language varieties and understanding the evidential status of genealogical groupings. Stable identifiers also help distinguish similarly named records.

For students, keep three separate notebook columns: structural value, genealogical classification and geographical information. Leave cells unresolved when evidence is missing. This small design choice prevents an attractive map from silently becoming a family tree. It also supports more careful reading of etymology, where historical claims require their own evidence.

Previous chapter · Contents · Next chapter

3. Define the clause before coding word order

Word-order labels are useful only after the target construction has been specified. Consider our original English example: “The gardener lifted the basket.” In this active clause, the gardener is the actor, lifted is the verb, and the basket is the affected object. The three major elements occur in subject–verb–object order, conventionally abbreviated SVO.

The sentence “The basket was lifted by the gardener” describes a closely related event through another construction. Treating its first noun phrase as the actor would confuse grammatical organisation with event roles. Translation equivalence does not make the two clause structures identical.

Matthew Dryer’s WALS chapter on subject, object and verb order sets out the relevant scope and explains dominant-order categories. It distinguishes a dominant order from the possibility of variation. A label summarises a defined comparison; it is not a command that every sentence must follow.

For an introductory exercise, restrict the records to active declarative clauses with two expressed full noun phrases and one lexical verb. State that this is a simplified teaching scope. Put questions, passive clauses and records with omitted participants in separate folders.

The semantic roles article helps readers identify event participants before deciding how a particular grammar expresses them. That order of work makes later classification much easier to audit.

Previous chapter · Contents · Next chapter

4. Work out an invented pattern without importing English

Here is a deliberately small artificial system. In the supplied key, nari means gardener, poku means basket, and tem means lift. The marker -ka identifies the actor; -mu identifies the affected participant. These meanings are stipulated for the exercise, rather than discovered facts about an actual language.

Record one is “nari-ka poku-mu tem”, translated as “The gardener lifts the basket.” Record two is “poku-ka nari-mu tem”, translated as “The basket lifts the gardener.” The second event is unusual, but its grammar is useful: it reverses the roles without changing the ordering template.

Both records place the actor phrase first, the affected phrase second and the verb last. Within this artificial dataset, they support an SOV pattern for the chosen construction. They also show that the endings carry role information. A reader who relies only on what normally happens in the world will misread record two.

Next, provide “nari-ka tem poku-mu” without a judgement about whether it is acceptable. The endings still identify the roles under the stipulated key, but the record alone does not establish whether this order is normal, marked or unacceptable. That requires further evidence.

Ask learners to separate three answers: what the words mean, which roles the markers identify, and what order occurs. An answer can be confident about one and uncertain about another. This separation is the central skill of the exercise.

Previous chapter · Contents · Next chapter

5. Read a gloss as an explicit analysis

An interlinear gloss shows how an analyst connects pieces of an expression with proposed meanings or grammatical functions. A free translation offers a readable rendering of the whole utterance. Keeping both allows a reader to inspect a proposal that a smooth English sentence might conceal.

Using the invented record from the previous chapter, the segmented form is “nari-ka poku-mu tem”. A compact gloss is “gardener-ACTOR basket-PATIENT lift”. A separate free translation is “The gardener lifts the basket.” The labels ACTOR and PATIENT are defined locally for this exercise; they should not be presented as universal replacements for every grammar’s terminology.

The Leipzig Glossing Rules provide conventions for aligning examples and glosses, displaying morpheme boundaries and explaining category labels. Crucially, they treat glossing as analysis. The observed expression and the analyst’s segmentation are related but distinguishable records.

A practical reading routine is to locate the original example, identify its boundaries, decode the abbreviations and then compare the free translation. If one element carries several proposed functions, a simple one-word English substitution may be inadequate.

When revising an analysis, preserve the source form and document what changed in the gloss. Otherwise a later reader cannot tell whether the language evidence changed or the interpretation changed. The connected morphology guide gives beginners a route into analysing word parts before working with unfamiliar conventions.

Previous chapter · Contents · Next chapter

6. Compare grammatical packaging feature by feature

Words and grammatical meanings do not necessarily line up one for one. A comparison should ask what a particular construction expresses and how its expression is distributed. Counting visible spaces is a starting observation about a written record, not a complete analysis of grammatical structure.

Imagine an artificial system in which “lum” means boat and “lum-ti” means boats. A second system uses “vek” for boat and “vek na” for boats in the supplied examples. Both express the contrast needed by this exercise, but one record uses an attached element while the other uses a separate one.

Before generalising, ask whether the pattern applies to every relevant noun, whether number must be expressed, and whether the same element has other functions. The exercise stipulates only the examples given. It does not establish a complete plural system.

The same discipline applies when familiar descriptions use broad labels such as analytic or agglutinative. Treat a broad label as an invitation to inspect the constructions underneath it. One label should not replace a record of which feature was measured and which exceptions were found.

A useful student comparison sheet has rows for number, tense, possession and participant marking. Each row can hold a different answer. The result is a profile rather than a single box into which a whole language must be squeezed. This helps learners see why a short translation can require several grammatical decisions without implying that one language is more sophisticated.

Previous chapter · Contents · Next chapter

7. Do not infer sounds from a writing system

Letters, speech sounds and contrastive sound categories belong to different levels of description. A project that compares writing systems may count written symbols. A project that compares sound systems needs relevant phonetic and phonological evidence. Substituting one count for the other changes the research question.

Suppose two invented writing conventions represent the same sound using “sh” and “š”. The first convention has two characters where the second has one. That difference alone says nothing about the number of sound contrasts in the language. A spelling count cannot settle an inventory question.

The International Phonetic Association’s chart supplies a standard reference for phonetic symbols. Using a recognised transcription system helps make a description readable, but a symbol still needs a documented interpretation in the particular source. A transcription is not automatically a full account of the sound system.

For classroom work, put orthographic form, transcription and meaning in separate columns. If only the written form is available, mark the sound comparison as unresolved. Do not ask a learner to infer an unfamiliar pronunciation from English spelling habits.

Continue into Writing Systems for the visual representation question and the pronunciation, IPA and transliteration guide for the distinction between preserving sound, changing script and translating meaning. Those are complementary tasks with different evidence requirements.

Previous chapter · Contents · Next chapter

8. Build categories that allow uncertainty

A coding sheet should explain both the positive and negative values of a feature. “Present” needs a defined test. “Absent” needs sufficient relevant evidence. “Unknown” records a limit in what can be established. These values support different conclusions and must remain distinguishable.

Consider a fictional grammar sketch that supplies six affirmative clauses and no discussion of questions. For a feature about question formation, the sketch may leave the answer unknown. It does not establish that the language lacks questions or lacks the construction being investigated.

Likewise, not applicable can be useful when the feature’s prerequisite is absent under the chosen definition. Its precise use needs a coding rule. A team should never let one coder use it for missing evidence while another uses it for a genuinely inapplicable comparison.

Create a short codebook with the feature question, allowed values, evidence required and examples of difficult cases. Add a field for source location and a note explaining any unresolved issue. A later reviewer should be able to reconstruct the decision without guessing what a colour meant.

Grambank is a major comparative grammatical resource; its documentation provides a route into explicit feature definitions and coding procedures. Beginners can learn from that discipline without pretending their small exercise is a comprehensive database. The immediate aim is a record whose uncertainty remains visible when it is sorted, counted or shared.

Add a worked decision to the codebook. For a fictional feature asking whether a dedicated plural marker occurs in ordinary noun phrases, list the relevant source examples and the scope of the claim. If the source explicitly describes its restricted use, preserve that restriction in a note. If it never addresses the construction, leave the value unknown. A categorical column is a convenient summary, so it needs an accompanying explanation whenever a binary answer would hide a material qualification. The note allows a later project with a different question to reassess the evidence rather than inheriting a misleading certainty.

Previous chapter · Contents · Next chapter

9. Check agreement between coders before counting

Two careful readers can disagree because the source is ambiguous, the categories overlap or an instruction is unclear. Recording disagreement is useful evidence about the method. Erasing it before discussion makes the final sheet look more reliable than the process supports.

In a fictional pilot, two students code twenty records independently. They agree on sixteen and disagree on four. Their raw agreement is 16 divided by 20, or 80%. This describes their matching decisions on that pilot; it does not prove that either student is correct.

Now inspect the four disagreements. Perhaps three involve clauses with omitted objects, while one involves a damaged transcript. The repair is specific: tighten the inclusion rule for the target clause and create an uncertainty note for the damaged record. Merely reminding both students to be careful does not address the causes.

Retain the original decisions, the revised rule and the adjudicated result. Then use a fresh set of records to see whether the clarification helps. Agreement on the same examples after a joint discussion is useful for training, but it is not an independent replication of the pilot.

For a student project, this modest procedure is often more valuable than a complicated reliability statistic used without understanding its assumptions. It teaches how a method improves. The comparison becomes stronger because another reader can see where judgement entered and what was done about it.

Previous chapter · Contents · Next chapter

10. State which languages and observations were sampled

A table of language varieties and a collection of utterances have different units. Fifty sentences from one variety are fifty sentence records, not fifty languages. A frequency measured within a text collection cannot be relabelled as the percentage of the world’s languages.

Consider a fictional selection containing ten varieties. Eight belong to Family A, while two belong to unrelated Families B and C. If all eight A varieties share a feature, an overall count of eight out of ten may chiefly reflect the heavy representation of A. It does not supply ten independent historical experiments.

A different sample might deliberately include fewer close relatives or provide separate summaries by family and region. There is no single sampling design that answers every question. A study describing a particular community’s texts should not discard those texts merely to resemble a worldwide survey.

Specify the target population first: the varieties in one archive, documented varieties in a selected area, or a broader research population. Then explain selection, exclusions and known gaps. Acknowledge where documentation availability shapes inclusion.

The WALS introduction discusses sample design and diversity. Use it as a reminder to inspect the composition behind a map. An honest small sample can answer a modest question well; an unspecified large sample can leave the central question unanswered.

Also separate sampling from coverage within each source. A grammar can be extensive while giving little space to the exact construction you need. A short paper can discuss that construction in detail while saying almost nothing about other features. Record the reason a source is suitable for this particular question. For an archive project, a coverage matrix can show which varieties have usable evidence for each feature. The resulting pattern of gaps may explain why two analyses of the same archive use different denominators. That difference should be explained before anyone interprets their percentages as contradictory findings.

Previous chapter · Contents · Next chapter

11. Read a percentage with its denominator attached

Here is an original numerical exercise. A fictional spreadsheet contains twelve varieties. Seven have feature X, three have a documented absence of X, and two are unknown. Among the ten varieties with known values, the observed share with X is 7 divided by 10, or 70%.

Among all twelve spreadsheet entries, seven are confirmed positive: 7 divided by 12, approximately 58.3%. This second fraction answers a different question. It is the confirmed-positive share of the full sheet, not an estimate obtained by treating unknown cases as negative.

For a simple sensitivity check, suppose both unknown entries eventually receive positive values. The full-sheet share would become 9 divided by 12, or 75%. If both became negative, it would remain 7 divided by 12. Thus 58.3% to 75% describes the range permitted by those unresolved cells for this sheet. It is not a statistical confidence interval or a worldwide prevalence estimate.

Now change the question to the percentage of speakers using varieties with X. The spreadsheet lacks speaker counts and a rule for multilingual people. The language-entry percentage cannot answer it. Equal weighting of entries is a methodological choice, not a natural measure of all human language use.

When reporting results, write the numerator, denominator and treatment of unknown values in the same sentence. That makes a calculation portable: readers can reuse the reasoning while seeing exactly where its interpretation stops.

Previous chapter · Contents · Next chapter

12. Test a relationship before calling it a rule

Features can be compared with each other, but a co-occurrence table needs careful reading. Suppose a fictional twelve-variety dataset records features X and Y for every entry. Of eight X varieties, six also have Y. Of four non-X varieties, one has Y.

The within-group proportions are 6 divided by 8, or 75%, and 1 divided by 4, or 25%. In this exercise, Y is more frequent among the X entries. That is a descriptive relationship in the table. It neither demonstrates that X causes Y nor establishes that every X variety has Y.

The two X entries without Y are especially useful. They show immediately that “all X implies Y” is false for this dataset, assuming the coding is sound. A tendency can remain worth investigating after a universal claim fails.

Before taking the pattern further, check how the varieties were selected, whether close relatives dominate a cell, and whether different definitions would change the assignments. A small conditional table can be sensitive to one corrected record. Preserve the original data and make any correction traceable.

The Logic root provides a natural continuation for distinguishing a universal implication, a counterexample and a probabilistic tendency. The typological skill is to connect those forms of reasoning to real coding decisions rather than treating a promising association as an explanation already secured.

Previous chapter · Contents · Next chapter

13. Transfer the method into learning and translation

Typological comparison can help a language learner ask better questions. If a familiar language uses position to distinguish participants, a learner can look deliberately for the cues used in a new construction. Those cues must still be learned from actual examples and reliable descriptions.

Use the invented nari system again. In “poku-ka nari-mu tem”, the endings identify the basket as actor and the gardener as affected participant under the supplied rules. A translation that makes the gardener lift the basket is more plausible in everyday life but wrong for the exercise. Faithfulness sometimes requires preserving an unusual event.

Now remove the markers from an incomplete record. A good reader notices that the evidence has changed and asks for context or a better source. Applying the earlier answer automatically would demonstrate memory of the example rather than transfer of the method.

For teaching, give learners a new vocabulary key and a different order pattern after a delay. Ask them to explain what evidence identifies the roles. Judge the explanation alongside the translation. This reveals whether the learner can separate form, meaning and uncertainty without help from familiar words.

The word-for-word translation article continues the practical problem. Typological awareness helps formulate questions; each particular translation still requires its own contextual and linguistic checks.

A second transfer task can change the audience rather than the grammar. Ask a learner to explain the same record first to a classmate who knows the marker key and then to a parent who has never seen it. The first explanation may use compact labels. The second needs the key, an example and a statement of uncertainty in ordinary language. Both should preserve the same analysis. This tests whether the learner understands the reasoning well enough to communicate it without relying on specialist shorthand. Clear explanation is part of a usable comparison, especially when its results travel beyond a language classroom.

Previous chapter · Contents · Next chapter

14. Repair misconceptions with counterexamples

Does an SOV label mean every utterance has three words in that order?

No. The label concerns a defined relationship among major elements in selected constructions. A sentence may contain several words inside one phrase, and a language description may discuss other permitted arrangements.

Does flexible order mean that arrangement carries no information?

No. Before calling two orders interchangeable, inspect the contexts and the source’s account of their use. The absence of a single rigid template does not remove the need for a grammatical description.

Does a shared feature prove that two languages are related?

No. Historical relationship requires historical evidence. A structural match can guide a research question without answering it. Similarly, a map point cannot describe every speaker, register or period associated with a language name.

Does a missing marker mean a language cannot express the relevant meaning?

That conclusion exceeds the observation. The information might be supplied elsewhere, inferred in context or expressed through another construction. Identify the comparison target before deciding what a negative coding means.

Can a classroom dataset establish a worldwide universal?

It can test the proposed claim against its own records and may contain a counterexample. Positive coverage in a small exercise does not settle a universal claim about unexamined languages. Keep the useful result: learners have practised defining scope and checking evidence. That achievement is more substantial than attaching a grand conclusion to a tiny table.

Previous chapter · Contents · Next chapter

15. Make the comparison usable by the next reader

A finished comparison should leave a clear trail from question to conclusion. Include the feature definition, variety identifiers, source locations, coding decisions, unresolved cases and counting rule. Keep the table and the explanatory prose consistent when corrections are made.

A compact final task is to exchange notebooks with another learner. Can that reader reproduce the count, identify why a record was excluded and locate the evidence for a disputed classification? If the answer is no, revise the record before adding another chart. Reproducibility begins with the information another person actually needs.

Use the directory below according to the next question. Linguistics supplies the wider analytic framework. World Languages provides broader orientation. Morphology and semantic roles support close analysis. Corpus work helps examine selected collections of usage. Translation and multilingual research show where differences in form matter in practical communication.

For further study, begin with a single documented feature in WALS or Grambank. Read its definition before viewing the distribution, inspect the supporting references and record the dataset version used. Do not combine columns from different resources until identifiers and definitions have been checked.

The gold standard is a comparison that remains useful after the original author leaves the room. Another reader should understand the categories, follow the evidence, recognise the limits and apply the method to a new case. That is how a subject label becomes a transferable intellectual habit.

Previous chapter · Contents · Explore the article directory

Sources and teaching-example notes

Primary resources checked on 21 October 2026. Read each resource’s feature definitions and citation instructions before reusing its data. This article does not calculate global frequencies from those databases.

• WALS introduction — Bernard Comrie, Matthew S. Dryer, David Gil and Martin Haspelmath
• Order of Subject, Object and Verb — Matthew S. Dryer, WALS Chapter 81
• Leipzig Glossing Rules — Max Planck Institute and University of Leipzig
• Grambank — comparative grammatical features and documentation
• Glottolog — languoid identification and classification principles
• International Phonetic Association — official IPA chart

All artificial forms and numerical records in the worked exercises were created for this guide. They are separate from the real-language documentation in the linked resources. Dataset versions and coding rules should be recorded for any independent analysis.

Return to reading routes · Education: learning transfers into life