This chapter discusses the morphological coding of evidentiality, which marks the source of information the speaker has for his or her statement. This chapter complements Chapter 77, which deals with the semantic distinctions of evidentiality. As was the case in the previous chapter, only grammaticalized evidentials are included here.
It turns out that evidentiality is marked across languages in a wide variety of ways. The following morphological means for encoding evidentiality are represented on the map:
|Go to map|
|No grammatical evidentials||181|
|Verbal affix or clitic||131|
|Part of the tense system||24|
These diverse means of coding evidentiality are a direct reflection of the origins of the evidentials in the respective languages. Thus, for instance, the fact that in some languages evidentiality is part of the verbal system means that these evidentials were originally tense morphemes. The same is true for the other ways of encoding evidentiality.
We turn now to a discussion of the different ways evidentiality is encoded in the sample.
From the accompanying map it appears that expressing evidentiality as a verbal affix or clitic is the most common strategy. With the exception of Africa it occurs on every continent. Example (1) is from Kannada (Dravidian; Sridhar 1990: 3), where the quotative morpheme –ante is attached to the negative verb. In Lezgian (Nakh-Daghestanian; eastern Caucasus; Haspelmath 1993: 148) the quotative morpheme –lda is attached to the main verb, as in example (2).
‘(It is said that) your book is not with him.’
‘They say that there will be a meeting today.’
In some cases the evidential morpheme is a clitic rather than an affix. In a number of languages the evidential can be attached to other word classes besides the verb. An example of such a language is Takelma (Takelman; Oregon; Sapir 1922b: 291-292). The evidential morpheme -ihiʔ, which functions as a Quotative, can be attached to any word class. This is shown in (3):
‘he said, it is said’
‘now, it is said’
In a number of languages the direct–indirect evidential distinction (these terms were defined in Chapter 77) is part of the verbal system. An example is shown in (4) from Turkish (Aksu-Koç and Slobin 1986), where there are two past tenses that can be used for evidential distinctions.
‘Ahmet must have come.’
Most languages that use the verbal system to denote evidential distinctions do so only in the past tense. Some languages, such as the Caucasian languages Mingrelian, Svan, and Tsova-Tush, have evidential distinctions in the present and future as well.
While in some languages the distinction between direct and indirect evidentiality has been grammaticalized (Turkish is such a language), this is not universally the case. In Georgian (Kartvelian), past tense indirect evidentiality has been grammaticalized as one meaning of the Perfect, but the corresponding Aorist past has not (yet) been formalized as a marker of direct evidentiality. An example is (5) (Boeder 2000: 285-286):
‘It has snowed.’
‘It has snowed.’
When a language uses separate particles for evidentiality, this is very strongly correlated with coding indirect evidentiality only. Whenever a language uses separate particles, it will only use them for indirect evidentiality. An example is (6), from Dumi (Tibeto-Burman; van Driem 1991: 263), which shows the use of a quotative particle ʔe:
‘What did he/they/you say he was eating?’
The only possible exceptions in the sample to the generalization that particles are only used for indirect evidentiality are Apalai (Koehn and Koehn 1986: 119) and Lega (Bantu; Botne 1995). In Apalai, the particle puh(ko) is used to denote visual evidence. The example given is shown in (7):
‘I could tell it was all gone.’
It is not clear, however, that this is a direct evidential, since from the translation it would appear that we are dealing with visual evidence after the fact, i.e., an inferential.
‘[It seems that] you may drink this beer.’
‘She is assuredly pounding rice [I can hear it].’
The origin of the direct evidential in Lega is a proximate pronoun, which explains the evidential’s status as a particle. Deictic elements frequently serve as source material for evidentials (see de Haan 2001).
There are several instances of evidentiality coded by means of a modal morpheme. In many instances this element is a separate modal verb, as in (9) from Dutch, where the modal verb moeten ‘must’ can encode indirect evidentiality:
Het moet een goede film zijn.
‘It is said to be a good film/ It appears to be a good film.’
In some languages the irrealis or subjunctive morpheme serves as an (indirect) evidential, as is the case in a number of Australian languages. Example (10) is from Gooniyandi (McGregor 1990: 550, Bill McGregor, p.c.), where the past subjunctive morpheme –ja can be an indirect evidential.
‘They were eating here not long ago (there is evidence…).’
‘They supposedly shot from Najig right up to Gumja.’
Languages which have more than one way of encoding evidentiality usually have a combination of a separate particle and a verbal affix. An example is Diyari (Australian; Austin 1981: 173), which has a Quotative particle pinti and an affix –ku which marks sensory evidence (hence a direct evidential):
‘They say he is coming.’
‘It looks/feels/smells like rain will come.’
The distribution of languages with and without evidentials was discussed in Chapter 77. This section focuses on the distribution of the different formal strategies for encoding evidentiality.
The distribution of some morphological markers appears to have a geographical connection. The encoding of evidentiality in the tense system is found most often in two areas often linked to areal studies, namely the Balkans and the Caucasus. The encoding of evidentiality is a prominent feature in most Turkic languages (see Johanson 2000) and also in several Caucasian families (e.g. in Kartvelian).
The evidential use of modals is mainly a western European feature. It occurs in most Germanic languages, as well as in Finnish. In these languages evidentiality is another interpretation of modal verbs. This means of encoding occurs occasionally elsewhere, usually as part of irrealis (or subjunctive) marking (as in Australian languages such as Gooniyandi).
In the languages of the Americas, evidentiality is most often encoded either as a verbal affix or as a separate particle. In certain language families (e.g., Eastern Tucanoan) it is part of the tense system.
In the other areas there is little or no areal patterning discernible. It would appear that in Asia affixation on the verb is more common than any of the other means, but this is by no means a fixed rule. Whether or not areal diffusion is wholly or partially responsible is still an open question.
It has been claimed that evidentiality can be considered an areal feature (see Haarmann 1970, Aikhenvald and Dixon 1998, and Johanson and Utas 2000, among others). This claim is probably correct, given the observed clusterings of features, both semantic and morphological. From the data it seems that languages in the same geographical area can adopt structurally similar evidential notions. This means that evidentiality is a transparent category, with respect to both its semantics and its morphological coding. Evidentiality is a category that diffuses easily from one language to another, even when these languages are genetically unrelated. Of course, the fact that this can occur is no guarantee that it will occur.