The parameter
Evidentiality is the grammatical marking of the source of a statement: whether the speaker saw the event, inferred it from its results, or was told about it. Every language can say how it knows something. The parameter asks whether the grammar obliges the speaker to say so, on the verb, in every declarative clause.
That distinction is easy to lose. English has ample lexical resources for source: "apparently", "reportedly", "is said to have". None is compulsory, and an English past tense makes no claim about evidence. A Turkish speaker choosing between -DI and -mIş cannot decline to choose. Aikhenvald reserves the term evidentiality for the grammaticalized case and calls the English resources evidential strategies, a line worth keeping because it names what is missing on the English side.
Evidentiality is also not epistemic modality, though the two are often conflated. A reported form says where the information came from, not how likely it is, and rendering it as "may have" swaps a claim about source for a claim about probability.
How languages vary
- Grammaticalized evidentiality: Turkic, with a two-way witnessed against non-witnessed contrast; Bulgarian and Macedonian, with renarrative forms built on the past participle; Cuzco Quechua, with the enclitics -mi for direct evidence, -si for report and -chá for conjecture; Aymara and many Amazonian languages with three or more terms.
- Lexical evidential strategies: English, French, Spanish, Dutch. Adverbs, reporting verbs and modals do the work, always optionally. German sits at the edge: its reportative subjunctive is close to a grammatical device in journalistic and legal register.
- Hearsay particle only: a single reported marker and no other evidential term, the smallest system Aikhenvald recognizes.
De Haan's WALS chapter 77 samples 418 languages: 181 have no grammatical evidentials, 166 have indirect evidentials only, and 71 mark both direct and indirect. His chapter 78 shows the marking is most often a verbal affix or clitic, in 131 of the same 418. Aikhenvald's own estimate, about a quarter of the world's languages, is lower than the WALS proportion; the two should not be reconciled by arithmetic, since the WALS sample documents the category rather than sampling proportionally.
Consequences for translation
Out of an evidential language into English the operation is a level shift, and it is compulsory: no English verb form carries the value. The only question is whether the translator performs it or lets the category lapse.
A Turkish reported past rendered as a plain English past tense silently deletes a grammatical claim about the source of the information. Nothing in the target marks the deletion. The English sentence is fluent, the reviser reading only the target sees no defect, and the reader is given as directly known something the speaker marked as second hand. In a witness statement, a police report or an asylum interview record that is not a stylistic matter: it is the difference between what the person saw and what the person was told, and it is frequently the point at issue. The professional response is explicitation under a stated policy: render the category where it is contrastive, record where it was allowed to lapse, and note it where an adverb cannot carry the load.
Two practical points follow. Explicitation accumulates: a suffix that recurs on every verb becomes an adverb on every English verb, and the text reads as relentlessly hedged, a register change the source did not make. And the reverse direction forces a commitment rather than losing one: translating an English past into Turkish or Bulgarian requires a form stating how the writer knows, which the source usually does not.
Examples
TR>EN
- Source: Ali parayı almış. / Ali parayı aldı.
- Target: Ali took the money, apparently. / Ali took the money.
- Comment: One suffix apart. The first says the speaker did not witness the taking; the second says the speaker vouches for it directly. Rendering both as "Ali took the money" merges the distinction a statement of evidence exists to record.
BG>EN
- Source: Кучето изяде рибата. / Кучето изяло рибата.
- Target: The dog ate the fish. / The dog ate the fish, I am told.
- Comment: The renarrative is built on the past active participle and drops the auxiliary in the third person. It is a full paradigm, not a particle, so it recurs in every clause of a reported narrative.
EN>TR
- Source: The suspect left the building at 22:00.
- Target: Şüpheli binadan saat 22.00'de ayrıldı. / ... ayrılmış.
- Comment: English states the fact without saying how it is known, and Turkish has no form that does the same. The translator must choose a witnessed or a reported past, and in a police document that choice attributes or withholds direct observation. Query it.
References
- Aikhenvald, A. Y. (2004). Evidentiality. Oxford: Oxford University Press.
- de Haan, F. (2013). "Semantic Distinctions of Evidentiality." In M. S. Dryer & M. Haspelmath (eds), The World Atlas of Language Structures Online, chapter 77. Leipzig: Max Planck Institute for Evolutionary Anthropology.
- de Haan, F. (2013). "Coding of Evidentiality." In WALS Online, chapter 78.
- Chafe, W. & Nichols, J. (eds) (1986). Evidentiality: The Linguistic Coding of Epistemology. Advances in Discourse Processes 20. Norwood, NJ: Ablex. Including Aksu-Koç, A. & Slobin, D. I. on Turkish and Friedman, V. A. on Bulgarian, Macedonian and Albanian.
- Johanson, L. & Utas, B. (eds) (2000). Evidentials: Turkic, Iranian and Neighbouring Languages. Berlin: Mouton de Gruyter.
- Faller, M. (2002). Semantics and Pragmatics of Evidentials in Cuzco Quechua. PhD dissertation, Stanford University.