Chapter Semantic Distinctions of Evidentiality

by Ferdinand de Haan

1. Defining the values

This map and the one following it show the presence of grammatical markers of evidentiality. On this map the semantic coding of evidentiality is represented, while Map 78A deals with the morphological coding of evidentiality.

Values of Map 77A. Semantic Distinctions of Evidentiality
No grammatical evidentials 181
Only indirect evidentials 166
Both direct and indirect evidentials 71

Markers of evidentiality express the evidence a speaker has for his/her statement. For the purposes of this map and the following one, evidentiality is divided into two broad sub-categories, direct evidence and indirect evidence. Only grammaticalized evidentials have been included in this study; this excludes lexical adverbs, such as reportedly  in John is reportedly ill, and biclausal constructions, such as It is said that John is ill. Auxiliary verbs and particles have been included. Also, a distinction is made between evidentiality, which is used only for the evidence for the statement, and epistemic modality, which marks the speaker’s degree of confidence in the statement (see Chapter 75). Although it is possible for evidentials to develop from epistemic modals (see Chapter 78), this is not common.

Direct evidentials are used when the speaker has some sort of sensory evidence for the action or event he/she is describing. Normally, a direct evidential denotes visual evidence. Example (1), from Fasu (Trans-New Guinea; Papua New Guinea; Loeweke and May 1980: 71), shows a visual evidential in the form of the circumfix a-…-re :

(1) Fasu



‘[I see] it coming.’

Other sensory information can be coded in evidentials as well. Example (2), from Koasati (Muskogean; Louisiana and Texas; Kimball 1991: 207), shows an auditory evidential, which expresses the idea that the action described, in this case the charring of meat, was perceived by hearing.

(2) Koasati





‘It sounds like the meat is charring.’

The other senses are rarely expressed individually. If they are expressed with an evidential, they are usually combined with auditory evidence to form a non-visual sensory evidential. Example (3), from Tuyuca (Eastern Tucanoan; Colombia and Brazil; Barnes 1984: 260), shows a non-visual sensory evidential. In Tuyuca, evidential affixes also express tense (past and present, but not future), person, number and gender. Example (3a) shows the auditory use of this evidential, while (3b) shows an example of one of the other senses (with a change in tense).

(3) Tuyuca








‘The motor roared.’ (I heard it)








‘My stomach hurts.’ (I feel it)

In the majority of cases, however, there is one general direct evidential, covering all modes of sensory evidence, although it is usually used most often for visual evidence. Example (4) comes from Kewa (New Guinea; Franklin 1971: 50). In this example, the most common interpretation of –ha  is that it denotes visual evidentiality, but other forms of direct evidence are possible.

(4) Kewa


‘He cooked it.’

Indirect evidentials are used when the speaker was not a witness to the event but learned of it after the fact. There are two broad sub-categories, inference and quotative. Inferential evidentials are used when the speaker draws an inference on the basis of available physical evidence. Example (5) comes from Khalkha (Mongolian; Street 1963: 129):

(5) Khalkha







‘He must have come.’

Quotatives (also known as reportatives, hearsay, or second-hand evidentials) are used when the speaker has been told about the action or event by another person. Example (6) is from Lezgian (Nakh-Daghestanian; eastern Caucasus; Haspelmath 1993: 148):

(6) Lezgian







‘They say that there will be a meeting today.’

As is the case with direct evidentials, more often than not both types of indirect evidence are grouped together into one general indirect evidential. This happens for instance in Dutch, where the verb moeten  ‘must’ (also used as an epistemic or deontic modal, like its English cognate verb must ) can be used for unspecified indirect evidence:

(7) Dutch

Het moet een goede film zijn.

‘It seems to be a good movie.’ (I have no direct evidence)

Unlike the Dutch verb moeten, English must  does not have an evidential sense. Whereas Dutch moeten  in its evidential sense has no epistemic modal interpretation, English must  does. The latter is therefore not an evidential in the way this term is used in this study (see de Haan 1999 for further details).

Many languages make a distinction of witnessed (direct) versus unwitnessed (indirect) actions in the past tense. These include Kartvelian languages like Georgian (Boeder 2000), many Turkic languages (see (8) below), as well as Komi-Zyrian (Finno-Ugric; Leinonen 2000), Haida (isolate; Alaska and British Columbia; Swanton 1911b), and Ika (Chibchan; Colombia; Frank 1990). Example (8) is from Turkish (Aksu-Koç and Slobin 1986), where a distinction is made between witnessed past (the morpheme -di ) and unwitnessed (-miş ).

(8) Turkish








‘Ahmet came.’

(witnessed by the speaker)








‘Ahmet came.’

(unwitnessed by the speaker)

Because evidentials are used to describe the speaker’s involvement with events, they tend to occur in realis contexts, especially in past tense situations (cf. example (8) above). Nevertheless, evidentials do occur in what can be described as irrealis situations. Example (9) from Barasano (Eastern Tucanoan; Colombia; Jones and Jones 1991: 116) shows an interrogative evidential, and (10), from Tsova-Tush (Nakh-Daghestanian; Georgia; Holisky and Gagua 1994: 180) shows an evidential with a future event.

(9) Barasano







‘Is he a toucan?’

(10) Tsova-Tush



‘S/he will be cutting it apparently.’

There are no languages in the sample that have grammaticalized direct evidentials, but no grammaticalized indirect evidentials.

2. Geographical distribution

Evidentials are common in the languages of the world, more common than usually assumed. They are found on every continent, but some very clear areal patterns emerge from the map: (1) evidentials are almost completely absent in Africa; (2) languages having only indirect evidentials are common in Europe; (3) languages with both direct and indirect evidentials appear to be clustered in the western United States, the western Amazon region, the Caucasus, and in the Himalayan and adjacent areas, but they are by no means restricted to these areas; (4) languages of the Americas are very likely to have at least indirect evidentials; (5) languages of the Pacific area, including New Guinea, are slightly more likely to have no evidentials. As a general statement, evidentiality would appear to be more of an areal feature than a genetic feature.

The most remarkable feature of the map is the almost complete absence of evidentials in Africa. The only languages with evidentials in the sample are Afrikaans, a Germanic language spoken in South Africa, Lega, a Bantu language in central Africa (which has both a direct and an indirect evidential according to Botne 1995), and Beja, a Cushitic language with a quotative construction.

The languages with indirect evidentials in Europe are mostly Germanic (though not English). Romance languages do not seem to have evidentials, with the exception of French, probably under the influence of Germanic. The same can be said for Finnish, which has evidentials derived from modal verbs, as do the Germanic languages. Evidentiality can possibly be considered to be part of the Balkan Sprachbund, even though languages such as Modern Greek and Romanian do not seem to have them. The presence of evidentiality in Bulgarian, Macedonian and Albanian may be due to Turkish influence.

In Asia, there are two areas with a high concentration of evidential systems: the Caucasus and the Himalayan area. In the languages of the Caucasus, especially the Kartvelian group, evidentiality is encoded with tense and aspect morphemes, although it can also be expressed with separate suffixes. On the South Asian subcontinent, the Indo-Aryan languages in general lack evidentials, except those that have been in contact with Tibeto-Burman or Iranian languages. In the southern part of the subcontinent, the Dravidian languages do have evidentials. The Tibeto-Burman family is rich in evidentials, although the phenomenon would seem to occur more frequently in the Tibetic group than elsewhere. Mon-Khmer languages and Thai languages do not seem to have evidentials.

In the Pacific region, evidentials are not found in the Polynesian languages or outside the Australia/New Guinea area, but otherwise are sporadically documented.

Evidentials are widespread in the Americas. In North America, the largest evidential systems can be found in California (especially the Pomoan languages) and the Southwest (in Western Apache and the Uto-Aztecan languages, but also in Kiowa-Tanoan). However, evidentials occur in almost every language family in the Americas. In South America, the largest evidential systems occur in the Vaupés River region in Western Brazil and adjacent areas.

3. Theoretical issues

Current research in evidentiality focuses on the interaction of evidentiality with other categories, such as epistemic modality (Palmer 1986, de Haan 1999), tense, and deixis (Floyd 1999, de Haan 2001). Also of interest is the role of evidentiality as a possible areal feature, given the ease with which it spreads from language to language (see Haarmann 1970, the papers in Johanson and Utas 2000, and Aikhenvald and Dixon 1998).