Skip to main content
Back to the atlas

Grammar & similarity

The typology lens

Colour the whole map by any single grammatical feature, or set two languages side by side and read exactly where they agree and differ. Grammar is coded for 203 of the 980 languages — the rest stay on the map, honestly greyed as “not grammatically profiled”.

203
languages profiled
of 980 total
431
features
195 Grambank · 186 WALS · 50 ext.
30,351
coded data points
value where recorded
9,499
agreement pairs
recorded Grambank agreement
  • Showing first 300 of 431 — refine the search.

Whether the pronoun system distinguishes inclusive ('we incl. you') from exclusive ('we excl. you').

Grambank GB028 · Pronoun

Coded for 138 of 203 grammatically-profiled languages (203 of 980 total) · 5 coded ‘?’ (unknown).

  • present103
  • absent35
  • coded ‘?’ (unknown)5
  • profiled, feature not coded60
  • not grammatically profiled
Loading map…
Loading grammar data…

Grambank’s ~195 variables are a standardized cross-linguistic baselinefor comparison, not the whole grammar of a language. Grey means the value is honestly unknown or not coded — never “absent”.

About this lens & its method

The data. Grammatical feature values come from Grambank v1.0.3 (~195 standardized variables) and WALS 2020.4 , with an Australianist extension layer — all CC-BY-4.0. They are joined to the atlas by Glottocode from a read-only build-time snapshot; nothing is queried live.

A baseline, not the grammar. Grambank’s variables are a cross-linguistic baselinefor comparison, not a complete description of any language’s grammar. A language coding few of them is not “simpler” — it is less fully profiled.

The similarity metric is grammar_recorded_agreement Similarity in recorded Grambank features”. It is the fraction of jointly-coded Grambankfeatures on which two languages carry the same value. ‘?’-unknown and N/A are excluded from both sides of that fraction. We always show n_joint — a 90% agreement over 8 shared features is not the same as over 150.

It is not relatedness. Recorded agreement is not overall grammatical similarity and not genetic relatedness — unrelated languages can share many baseline values. Unknown and not-coded cells render an explicit grey, never conflated with “absent”.

Data release 1.1.0 · reproducible via pnpm atlas:build-data· sources & licences in the methodology. A spreading language lineage is not the same as a moving population.