Korean Evidentials and Assertion

Sentences with the Korean suffix -te express evidentiality (Sohn 1994, Cinque 1999). For
example, (1) conveys the direct evidential meaning that the speaker witnessed the event, and thus
contrasts with the regular nonevidential sentence without -te in (2), which is not specified with the
evidential meaning. However, it has also been noted that sentences like (1) imply an abstract concept
regarding the speaker's attitude such as `psychological distance' and `weakened reliability' (Shin 1980) or
`irresponsibility' (Kim 1981). Such meanings are unexpected since direct evidentiality should convey that
the proposition in question is more reliable and trustworthy, and thus the speaker feels certain about it
(Willett 1988). How can the two aspects of meaning conveyed by -te sentences be reconciled? My claim
is that in Korean evidentials the speaker relates the speaker's association to the information without
believing or making a commitment to it. To capture this in formal terms, I propose that a Korean
evidential sentence (direct or indirect) does not express an assertive speech act but rather a presentative
speech act, in the sense of Faller (2002).
First, we can verify that Korean evidential sentences are not assertive. Clauses with -te are
allowed as complements of the verb malha `say' but not other attitude verbs, such as mit `believe' and
tanenha `assert', as illustrated in (3a)(3c). In contrast, non-evidential sentences have no such restrictions
(see (3d)). This is also true of the indirect evidential sentences with the epistemic modal suffix -keyss
denoting `possibility' (or `weak necessity'): the non-evidential (4a) is fine but the evidential (4b) is not.
This is because (4a) indicates the speaker's assertion about the probability of the proposition Mary
becomes the president, whereas in (4b), the speaker simply presents the probability of the proposition
based on his (her) reasoning without asserting it.
Second, the presentative nature of -te sentences is revealed by special conditions on the suffix -te.
-Te refers to the speaker's perceptual field in the past reference time and also has the restriction on the
speaker: the speaker should be a passive perceiver not an active participant of the event described by the
sentence. So when the speaker plays a voluntary agentive role of the event, the sentence is unacceptable,
as illustrated in (5)(6). In contrast, regular non-evidential sentences do not exhibit this restriction. When
the speaker utters a direct evidential sentence like (1), (s)he objectively conveys the proposition that refers
to a state of affairs that is perceived through his (her) senses. This indicates that the speaker of an
evidential sentence is not the person that is actively involved in making a judgment about the proposition,
but simply serves as a channel through which the proposition is obtained and delivered to the hearer. This
means that the speaker is neutral about his (her) attitude toward or belief in the proposition, which
explains why the direct evidential sentence (1) expresses those concepts mentioned above.
A final argument that evidentials are presentative speech acts comes from the fact that, unlike
non-evidential sentences, they always require the presence of the actual hearer and thus are rarely used in
written texts. In this respect, evidential sentences have a property of ditransitive predicates.
To capture the above facts, I propose the sincerity condition on the speech act of presentation in
(7), which says that the speaker s simply delivers to the hearer h the fact that (s)he has evidence v of
proposition p, and the evidence can be direct, inferential indirect, or reportative indirect. The condition
does not specify the speaker's belief in the proposition.
In sum, I argue that a presentative speech act is what crucially distinguishes evidential sentences
from non-evidential sentences in Korean and consequently the role of the speaker differs between the two
types of sentences. I suggest that evidential and non-evidential sentences differ in terms of the evaluation
world--while the latter use the actual world, the former use the speaker's perceptual world--and discuss
a further consequence in the epistemic modal interpretation. Furthermore, using Nuyts' (2001) notion of
speaker's subjective/objective epistemic evidence, I explain why evidential sentences, unlike non-
evidential sentences, are not assertive. I suggest that Korean speakers use evidential sentences to be
relieved of the burden of full responsibility for the claim made on his (her) subjective evidence only. One
implication of my analysis is that evidentials in general may lack assertive speech acts. On the other hand,
it is possible that the assertive mode is subsumed under the evidential system in some languages but not
others, thus capturing a fundamental difference in the use and meaning of evidentials across languages.

(1) mary-ka ku phyenci-lul ssu-te-la.
Mary-NOM that letter-ACC write-E.PAST-DEC
`[I saw] Mary was writing that letter.'
(2) mary-ka ku phyenci-lul ssu-ko iss-ess-ta.
Mary-NOM that letter-ACC write-PROG-PAST-DEC
`Mary was writing that letter.'
malha-yss-ta.
(3) a. mary-nun john-i ku pheynci-lul ssu-te-la-ko
Mary-TOP John-NOM that letter-ACC write-E.PAST-DEC-COMP say-PAST-DEC
`Mary said that [she saw] John was writing the letter.'
mit-ess-ta.
b. #mary-nun john-i ku pheynci-lul ssu-te-la-ko
Mary-TOP John-NOM that letter-ACC write-E.PAST-DEC-COMP believe-PAST-DEC
`Mary believed that John was writing the letter.'
tanenha-yss-ta.
c. #mary-nun john-i ku pheynci-lul ssu-te-la-ko
Mary-TOP John-NOM that letter-ACC write-E.PAST-DEC-COMP assert-PAST-DEC
`Mary asserted that John was was writing the letter.'
tanenha-yss-ta.
d. mary-nun john-i ku pheynci-lul ssu-ko iss-ess-ta-ko
John-NOM that letter-ACC write-PROG-PAST-DEC-COMP assert-PAST-DEC
Mary-TOP
`Mary asserted that John was was writing the letter.'
tanenh-yss-ta.
(4) a. mary-nun john-i hoycang-i toy-keyss-ta-ko
Mary-TOP John-NOM president-NOM become-MOD-DEC-COMP assert-PAST-DEC
`Mary asserted that John would become the president.'
tanenh-yss-ta.
b. #mary-nun john-i hoycang-i toy-keyss-te-la-ko
Mary-TOP John-NOM president-NOM become-MOD-E.PAST-DEC-COMP assert-PAST-DEC
`Mary asserted that John would become the president.'
(5) a. nay-ka/nun nwunmwul-i nao-te-la. -- Unaccusative
tear-NOM come.out-E.PAST-DEC
I-NOM/TOP
`[I noticed] I had tears coming to/out of my eyes.'
b.??/*nay-ka/nun wul-te-la. -- Unergative
I-NOM/TOP cry-E.PAST-DEC
`[I noticed] I was crying.'
(6) a. nay-ka/-nun phiano-soli-ka tul-li-te-la. -- Passive
piano-sound-NOM hear-PASSIVE-E.PAST-DEC
I-NOM/-TOP
`[I noticed] I heard a piano.
b.??/*na-ka/-nun phiano-soli-lul tut-te-la. -- Active
I-NOM/-TOP piano-sound-ACC hear-E.PAST-DEC
`[I noticed] I heard a piano.
h, v[ Have (s, v for p)])}
(7) PRESENT(p): SINCERITY CONDITION = {Give(s,
s = speaker, h = hearer, v = evidence, p = proposition
References
Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford:
Oxford University Press.
Faller, Martina. 2002. Semantics and Pragmatics of Evidentials in Cuzco Quechua, Stanford Univeristy:
PhD dissertation.
Kim, Young-Hee. 1981. Hoysanmwunuy inchingceyakkwa chaykimseng (Person Constraint on the
Retrospective Sentences and Responsibility). Hankwuhak (Korean Studies) 10:37-80.
Nuyts, Jan. 2001. Subjectivity as an Evidential Dimension in Epistemic Modal Expressions. Journal of
Pragmatics 33:383-400.
Sohn, Ho-Min. 1994. Korean. London: Routledge.