Aikhenvald (2004): Evidentiality typology #
@cite{aikhenvald-2004} @cite{de-haan-2013} @cite{barnes-1984} @cite{oswalt-1986}
Cross-linguistic typology of grammatical evidentiality, anchored on Aikhenvald's Evidentiality (Oxford University Press, 2004) and contrasted with @cite{de-haan-2013}'s WALS Ch 77/78 chapters.
This study file holds Aikhenvald's cross-linguistic generalisations on the
18-language sample. Per-language profiles live in
Fragments/{Lang}/Evidentiality.lean as evidentialityProfile : EvidentialityProfile — constructed via EvidentialityProfile.fromWALS so
WALS Chs 77/78 are auto-pulled and the editorial fields (markers, notes,
attestedEvidentials, family) are local per-language commitments.
Sample composition #
18 languages chosen to span:
- No-evidentials (English, Mandarin — 2 languages per Aikhenvald;
WALS additionally codes French, German, Japanese, Korean, Finnish as
indirectOnlyon a looser criterion). - Indirect-only (Georgian, West Greenlandic — per Aikhenvald; WALS codes Georgian as directAndIndirect).
- Direct + indirect (Turkish, Bulgarian, Tibetan, Abkhaz — per Aikhenvald; WALS agrees on Turkish/Bulgarian, codes Abkhaz as indirectOnly).
- Three-or-more (Quechua, Aymara, Tuyuca, Kashaya, Tariana — per Aikhenvald; WALS has no entry for Quechua/Aymara, lumps Tuyuca/Kashaya/Tariana into directAndIndirect).
Aikhenvald's typology distinguishes 4-term and 5-term systems (Tuyuca and
Tariana have 5 obligatory categories: visual / nonvisual / inferred /
assumed / reported). The local EvidentialSystem enum collapses these into
threeOrMore; the markers field records the finer-grained inventory.
Aikhenvald 2004 vs de Haan 2013 (WALS) divergences #
10 of 18 sample languages diverge between Aikhenvald's analysis and de Haan's WALS coding. The Studies file record-overrides per Aikhenvald and exposes divergences as first-class theorems in §6 — making the cross-author disagreement visible per linglib's interconnection-density thesis.
What this file deliberately omits #
Aggregate-count theorems (sample_system_counts, sample_coding_counts,
sample_size, etc.) trimmed per the no-aggregate-count anti-pattern memory
note. Substantive structural / implicational claims kept.
Per-language profiles drawn from Fragments/{Lang}/Evidentiality.lean
via EvidentialityProfile.fromWALS. 10 languages are record-overridden
where Aikhenvald 2004 disagrees with de Haan 2013 (WALS); see §6 for
first-class divergence theorems.
18-language Aikhenvald sample.
Equations
- One or more equations did not get rendered due to their size.
Instances For
Every language without grammatical evidentials has notApplicable coding.
Every language with grammatical evidentials has a non-trivial coding.
The system and coding fields are consistent: noGrammatical iff notApplicable.
Languages with three-or-more evidential choices always include a direct evidence category. Follows from the type definition; verified against the sample.
Every language with direct evidence has either two-choice or three-or-more (never indirect-only or no-evidentials).
The complexity hierarchy: noGrammatical < indirectOnly < directAndIndirect < threeOrMore.
Western European languages (English, French, German per Aikhenvald) all lack grammatical evidentials. The Balkan Sprachbund is the notable exception (covered by Bulgarian).
East Asian languages (Mandarin, Japanese, Korean per Aikhenvald) all lack grammatical evidentials. East Asia is an evidential-free zone.
Andean evidential cluster: Quechua and Aymara both have three-or-more systems coded as verbal affixes. Well-known areal feature.
Vaupés-Amazonian rich evidential systems: Tuyuca and Tariana, from different families but in contact in the Vaupés, both have three-or-more categories with five distinctions each.
Americas languages (Quechua, Aymara, Tuyuca, Kashaya, Tariana) all have three-or-more systems via verbal affixes. The Americas have the highest density of complex evidential systems.
TAM-fused evidentiality is characteristic of the Balkans and Caucasus. In the sample: Turkish, Bulgarian, Georgian, Abkhaz.
All languages with three-or-more systems use verbal affixes.
10 of 18 sample languages diverge between Aikhenvald's analysis and
de Haan's WALS coding. The Aikhenvald values are this Studies file's
record-overrides; the WALS values are the Fragment-side
Fragments.{Lang}.Evidentiality.evidentialityProfile (which goes
through EvidentialityProfile.fromWALS). Linglib's
interconnection-density thesis: incompatibilities visible.
Aikhenvald vs de Haan: French. Aikhenvald says no grammatical evidentials;
de Haan's WALS Ch 77 codes the journalistic-conditional reportative as
indirectOnly.
Aikhenvald vs de Haan: German. Aikhenvald excludes modal sollen/wollen from grammatical evidentials; WALS counts them.
Aikhenvald vs de Haan: Japanese. Aikhenvald treats soo da/rashii as
modal, not evidential; WALS classifies them as indirectOnly.
Aikhenvald vs de Haan: Korean. Aikhenvald excludes -deo- from grammatical evidentials.
Aikhenvald vs de Haan: Finnish. Aikhenvald excludes modal verbs from grammatical evidentials.
Aikhenvald vs de Haan: Georgian. Aikhenvald says indirect-only; WALS Ch 77 codes as direct-and-indirect.
Aikhenvald vs de Haan: Abkhaz. Aikhenvald says direct-and-indirect with TAM-fusion; WALS codes as indirect-only with verbal-affix coding.
Aikhenvald vs de Haan: Tuyuca. Aikhenvald distinguishes 5-term system
(visual/nonvisual/apparent/secondhand/assumed); WALS Ch 77 lumps into
the directAndIndirect 2-way bucket and Ch 78 codes as partOfTAM.
Aikhenvald vs de Haan: Kashaya. Aikhenvald distinguishes 4-or-5-way
system; WALS lumps into directAndIndirect.
Aikhenvald vs de Haan: Tariana. Aikhenvald distinguishes 5-term system
(Vaupés area); WALS lumps into directAndIndirect and codes as partOfTAM.