Korean and Mongolian have several grammatical forms that can be used to describe past situations. In Korean, -te- and -ess- are largely used for past situations, whereas Mongolian employs -laa, -jee and -v for past situations. This article discusses the grammatical properties of these suffixes, focusing on their semantics and subject restriction. Though the two suffixes in Korean can describe past situations, they differ in their semantic categories. While -te- is an evidential marker, -ess- is a tense marker. Mongolian suffixes belong to the same category of past tense, but they have different evidential connotation. While -laa implies the speaker’s firsthand knowledge about the situation, -jee indicates secondhand knowledge. The suffix -v is neutral in terms of evidential meaning. Some suffixes in Korean and Mongolian seem to show similar subject restriction. The Mongolian -jee and the Korean -te- do not usually occur with the first person subject. Both show the ‘first person effect’. When they are used in the context of a first person participant, they imply lack of control or awareness on behalf of the speaker. This article will, however, show that the subject restrictions observed with the two suffixes are different in their nature and are caused by different motivations. Based on our observation, we propose to distinguish firsthand evidentials into three subtypes; experiencer-oriented, performer-oriented and observer-oriented evidentials. We also revise the ‘first person effect’ to deal with examples of the Korean -te-.

Mongolian has three suffixes (-laa, -jee, -v) which can be used to describe past situations.2

As shown in the following example (4), the three suffixes can occur with a past time adverbial but not with a present or future time adverbial to describe an on-going present or future situation.

The three suffixes are all past tense markers in that they describe past situations and cannot be used to describe present or future situations.4 The three are, however, different in their evidential meaning. Examples (5–7) show their evidential differences.

Sentences (5–7) describe the same event of [the teacher’s coming here], casting it as a past situation. They are, however, different in the way that the speaker acquired information about the event. Sentence (5) with -lee implies that the information is firsthand. The speaker has direct knowledge of the event: he himself witnessed it. In sentence (6) with -jee, the information is non-firsthand. The speaker has not witnessed the event himself. Unlike sentences (5–6), sentence (7) does not say anything about how the speaker obtained the relevant information. It simply describes the fact as a past situation. Based on this observation, we can characterize the three past suffixes of Mongolian as showing evidential distinction.5

2Mongolian examples in this article are taken from Song (1997). Mongolian has another suffix -san which can describe past situations. However, it is not a tense marker but an aspect marker, Perfective, and does not indicate any evidential meaning, for which reason it is not included in this article (cf. Song 1997).  3The Mongolian Cyrillic letters are transliterated as /a b v g d ye yo j z i ĭ k l m n o ö p r s t u ü f x c č š šč ' ī ` e yu ya/, respectively. The following abbreviations are used: Acc (accusative), Aug (augmentative), Cl (classifier), Conn (connective), Dat (dative-locative), Decl (Declarative), Dim (diminutive), Eyewit (eyewitness), Fem (feminine), First (firsthand), Gen (genitive), IK (indirect knowledge), Imm (immediate), Impfv (imperfective), Infr (inference), Intr (intransitive), Loc (locative), Nom (nominative), Nonfirst (non-firsthand), Nonvis (non-visual), Past (past), Pfv (perfective), Pres (present), Refl (reflexive), Rem (remote), Seq (sequence), Sg (singular), SS (same subject marker), Sub (subject) Top (topic), Vert (vertical), Vis (visual).  4It is generally accepted in the literature that the three Mongolian suffixes are past tense markers (cf. Luvsanvandan 1968:75-78, Mishig 1978:119, Binnick 1990, Önörbayan 1994:203, Song 1997:160–210).  5The suffixes -laa and -jee are labeled as ‘Direct Knowledge Past’ and ‘Indirect Knowledge Past’ respectively in Song (1997). Along with other distinctions, the three suffixes were often differentiated in their remoteness in previous studies: Immediate Past -laa, Recent Past -v and Remote Past -jee in Mishig (1978:119): Very Close Past -laa, Simple Past -v and Relatively Close Past -jee in Ozawa (1963:45, 66–69). See Song (1997:160–183) for previous studies about the three suffixes and their problems.

Korean has two suffixes that can describe past situations, -te- and -ess-.6

In the literature, -ess- is mostly considered as past tense (Ko 2004:245–251, Sohn 1999:362, Moon 2009:125–160).7 It describes a situation which happened in the past. The suffix -te- is one of the most controversial grammatical forms in Korean. Among many grammatical categories attributed to this suffix such as past, past imperfective, and past perception, ‘retrospective’ has been the most favored category in the literature.8 Since Song (1998, 2002), however, firsthand evidential (past sensory observation) has been gaining support in recent research (Lim 2010, Lee 2011, Kwon 2012, Kim 2012). This article also argues that -te- is a firsthand evidential marker. While -ess- cannot be used to describe an on-going situation at present or a future situation, -te- is not confined to a past situation. It can also describe an on-going present or future situation.

Example (11) describes a present situation and example (12) a future situation. They disprove the possibility that -te- is a past tense marker. The semantic property that is shared in examples with -te- is that the information about the situation conveyed in the sentence is obtained through the speaker’s firsthand observation and the information is obtained before the speech time. Example (10) with -te- indicates that the speaker witnessed the situation of Jinwoo eating beef, whereas example (9) with -ess- does not say anything about how the speaker obtained the information. If the speaker does not have firsthand information about the situation, it is not appropriate to describe the situation with -te-. In other words, -te- is a firsthand evidential marker. Following Song (1998, 2002), this article argues that -te- is an evidential marker of past sensory observation.9 It indicates that the information is obtained directly by the speaker and that the temporal point of obtaining the information was before the speech time.

One may think that the Korean -te- and the Mongolian suffixes, especially the firsthand past -laa, are similar. The suffixes -te- and -laa have firsthand evidential meaning and are related with the temporal meaning of the past. They are different, however. While -laa is basically a past tense marker, -te- is a firsthand evidential marker. Their temporal meaning of the past is different in their nature. The suffix -laa indicates that the situation happened in the past. It places the situation before the speech time, hence it is a past tense. The suffix -te- does not overtly indicate when the situation happened. It only tells us that the relevant information about the situation was obtained in the past by the speaker himself, hence it is a firsthand evidential. Then the two suffixes of Korean can be characterized as follows.

Unlike the three suffixes in Mongolian, which are past tense markers having evidential connotations, the two suffixes in Korean, -ess- and -te- belong to different grammatical categories: -ess is a past tense marker and -te- a firsthand evidential marker.

6In this article, the transcription of Korean examples follows the Yale Romanization System.  7Some research analyzes the suffix -ess- as a perfective aspect marker (Nam 1978:12, Suh 1988). Whichever category it belongs to, it does not have any evidential meaning as its basic meaning.  8See Song (1998) for previous studies on -te-.  9Korean also has other evidentials such as present sensory observation -ney, indirect knowledge evidential -keyss- and reported evidential -tay, which will not be discussed in this article. See Song (2007, 2011) for details about them.

Evidentials often interact with the grammatical person of participants. Some evidentials are preferred to other evidentials in the first person context. Languages with grammatical evidentials normally choose a firsthand (direct knowledge) evidential to describe the speaker’s own action or situation (Aikhenvald 2003a: 16). Nambiquara (Macro-Tucanoan: Brazil) has a three-way opposition of evidentials: ‘observation’, ‘deduction’ and ‘narration’. First person forms are not permitted for ‘deduction’ or ‘narration’ (Lowe 1972: 373–4). Tuyuka (Tycanoun: Brazil and Columbia) demonstrates five evidential categories: ‘visual’, ‘nonvisual’, ‘apparent’, ‘secondhand’, and ‘assumed’. First person subject is used with either a visual or a nonvisual evidential but not with an apparent or assumed evidential (Barnes 1984: 261). In Eastern Pomo, which has four evidential forms of non-visual, logical inferential, hearsay/reportative and direct knowledge, the direct knowledge suffix -ya is much more frequent in first person constructions (McLendon: 2003:108).

Mongolian shows similar restrictions. While non-firsthand past -jee is not used with first person subject, firsthand past -laa or neutral past -v are allowed with first person (examples 14–15).

The reason firsthand evidentials are preferred in the first person context seems self-evident. Unless we are unconscious or forgot what we did, we perceive our own activities through firsthand information, not through secondary or hearsay information. It would be counterintuitive for the speaker to present their own activity or situation with non-firsthand evidentials. Therefore, firsthand evidentials are preferably employed to describe first person situations. Song (2007) argues that first person sentences carry such pragmatic presupposition and evidential implication.

Since the first person sentence implies firsthand evidential, employing nonfirsthand evidential for a first person situation would cause a semantic collision, which is the reason non-firsthand evidentials are avoided in first person contexts.

The semantic collision, however, does not always prevent the use of first person with non-firsthand evidential. Some languages allow non-firsthand evidential with first person, adding semantic overtones of lack of control, volition, intention or awareness with first person, the so-called ‘first person effect’ in Aikhenvald (2004: 220).

In Jarawara (a Madi language, southern Amazonia), if the speaker knew what he was doing when he got drunk (he deliberately got drunk), firsthand evidential is appropriate (example 17). If the speaker woke up drunk and didn’t remember what he had done the previous night, non-firsthand evidential is used (example 18) (Dixon 2003: 170, Aikhenvald 2004: 221).

When non-firsthand evidential, inferential, is used in a first person context in Yukaghir (a Yukaghir language: Saha Repbulic), it implies inadvertent actions over which one has hardly any control (Maslova 2003: 229, Aikhenvald 2004:221–222).

Mongolian also shows examples of first person effect. Though Mongolian usually does not allow -jee with a first person subject, there are exceptions to this subject restriction. Verbs like unt- ‘sleep’, sogt- ‘get drunk’, mart- ‘to forget’, and uxaan ald- ‘to lose consciousness’ allow the non-firsthand past suffix -jee with a first person subject, as exemplified in (21-23).

Examples (21–23) allow all the three past suffixes with a first person subject. Examples with -laa are used when the speaker had some sort of control or awareness of the situation, whereas those with -jee are used when the situation was out of the speaker’s control or awareness. The sentence with the suffix -laa in (21) implies that the speaker himself was aware of his sleeping: he might have wanted to sleep deeply and prepared for a good sleep. The one with the suffix -jee implies that the speaker was not aware of his sleeping at the event moment: he might have had a lot of work to do before he went to bed but fell asleep unconsciously.

In Aikhenvald (2004: 220), ‘first person effect’ is described as follows:

Aikhenvald (2004)’s wording of ‘first person effect’ can be rearranged as follows:

The definition of ‘first person effect’ in Aikhenvald (2004) and its rearranged one in (24) are, however, rather misleading. They seem to give the interpretation that the usage of first person in non-firsthand evidential raises the effect of granting such meanings. We argue that the process of first person effect is different from such assumption. Firsthand evidentials describe situations about which the speaker has firsthand information, whereas non-firsthand evidentials indicate that the information of the speaker is not firsthand. In this section it is argued that first person is not allowed with non-firsthand evidential due to semantic clashes between grammatically marked non-firsthand evidential and the pragmatically implied firsthand evidential meaning of a first person sentence. The situations of ‘first person effect’ are those where such a semantic clash is cleared or weakened because the first person participant is not in an ideal condition where they can have firsthand information about the situations. In other words, when a first person participant is not aware of, lacks control or does not intend what is happening, non-firsthand evidentials are allowed to describe their own situations. In this article, we retain the term ‘first person effect’ rather than trying to replace it with a new one, but the definition of ‘first person effect’ should be changed.

It is a well-known and baffling phenomenon in the literature of Korean linguistics that Korean -te- shows subject restriction similar to those we observed related to non-firsthand evidentials. As shown in examples (26 and 27), the suffix -te- is not normally used with a first person subject, while -ess- can be. The speaker cannot use -te- in connection with his own action.

On the surface, the subject restrictions of the Korean -te- (examples 26a and 27a) look similar to the subject restrictions we observed with the Mongolian -jee (examples 14b and 15b). However, there is a crucial difference between the two. Cross-linguistically it is non-firsthand evidentials that show subject restriction with first person. In Mongolian, it is the non-firsthand past -jee not the firsthand past -laa or the neutral past -v that shows the first person restriction. Firsthand evidentials are actually preferred in describing first person situations in languages like Mongolian. The suffix -te- in Korean is, however, not a non-firsthand but a firsthand evidential. It should be explained why, unlike in other languages, the firsthand evidential -te- is not allowed in the first person context in Korean.

Korean -te- also displays exceptions to the subject restriction. When the speaker was not aware of or lacked control or intention in relation to what was happening, the first person subject can be used with -te- (examples 28–30).

Situations in examples (28–30) look similar to those of ‘first person effect’ discussed above. The situation is uncontrolled or non-intentional, or the speaker was not aware of the situation when it was happening. They, however, differ from typical examples of first person effect. If we follow the criteria for first person effect set in (25), non-firsthand evidentials appear in the first person context when the first person participant does not have awareness, control, or intention in relation to the situation. Korean examples (28–30) are different in that they have the firsthand evidential -te- instead of a non-firsthand evidential.

Table (1) shows how the Korean -te- and the Mongolian -jee are similar but different. It should be explained why the Korean evidential -te- behaves differently from firsthand evidentials in other languages like Mongolian. In the following section we will explain what causes first person restriction in -te- and how we can explain the ‘first person effect’-like phenomenon in Korean.

Evidential systems across the world vary in how complex they are, and in what meanings they encode (Aikhenvald 2004:23). Evidentials labeled with the same term can encode different scopes of information in individual languages. The information scope encoded in firsthand evidential differs within and across languages. Some languages may have different firsthand evidentials depending on which sensory observation is used to acquire the information. Languages like Tariana (a North Arawak language spoken in the Vaupes) or Oksapmin (a TransNew Guinea language spoken in Papua New Guinea) differentiate visual information and non-visual information, both of which should be treated within firsthand evidential.

In Tariana, visual evidentials are used to express information obtained through seeing (example 31), whereas non-visual evidentials are used to describe things one can hear but not see (example 32) (Aikhenvald 2003b:137–8).

In Oksapmin, if the information was acquired visually, the verb is formally unmarked as in example (33). If the events were perceived through senses other than sight (hearing, tasting, smelling or feeling), they are expressed with a verb stem plus the verb ‘do’ (Aikhenvald 2004:46–47) as shown in example (34).

Firsthand evidentials can also be divided into subgroups according to the temporal point when the relevant information is acquired. Korean has two firsthand evidentials, -te- and -ney, and they differ in terms of their temporal point of acquiring information. While -te- indicates the information is acquired before the speech time, -ney denotes that the information is acquired at the speech moment (Song 2007, 2011).

In (35) the information was acquired in the past, whereas the temporal point when the information is acquired in (36) is concomitant with the speech moment.

In this article we propose that firsthand evidentials can also be divided into subtypes depending on the referential scope of the observee (the one/thing who/which is observed).10 What is described by firsthand evidential may differ according to what can be the object of the evidential origo’s observation. Typical examples of firsthand evidentials describe what the speaker experienced as an active or passive participant in a situation. The speaker may tell their autobiographical stories or someone else’s events or situations that they observed as a spectator. The speaker may be the performer of the event or just an observer of an event, which was performed by someone else. In this case, the referential scope of the observee includes the evidential origo (the speaker), the hearer and other people or things. This type of evidential does not show subject restriction. The subject can be first, second or third person. The Mongolian firsthand past -laa belongs to this type of firsthand evidential along with most firsthand evidentials reported in the literature. We call this type of firsthand evidential ‘experiencer-oriented evidential’.

There are, however, other kinds of firsthand evidentials whose referential scopes are different. In Kashaya (a Pomo language spoken in northern California), among other firsthand evidentials, they have ‘performative’ suffixes which ‘signify that the speaker knows of what he speaks because he is performing the act himself or has just performed it’ (Oswalt 1986:34–35).

What is special with performative suffixes in Kashaya is that the subject is always the first person. In other words, the performative suffixes cannot be used to describe an event performed by someone else except the speaker.12 When the speaker has firsthand evidence about someone else’s event, Kashaya employs different evidential forms such as factual or visual, which are exemplified in (39-40).

A similar firsthand evidential is reported in Central Pomo, which has a form used to mark ‘personal experience of one’s own actions’ (Mithun 1999:181). Central Pomo has various evidential enclitics, as shown in (41). As in Kashaya, Central Pomo employs different firsthand evidentials depending on who is the performer. When the performer is not the speaker, it uses the firsthand personal experience =ya (example 41b). When the performer is the speaker, it uses the personal agency =la (example 41f).

Performative in Kishaya and personal agency in Central Pomo are firsthand evidentials. They, however, differ from the ‘experiencer-oriented evidential’ in that they are exclusively used to describe events performed by the speaker himself. In this type, the referential scope of the observee is confined to the evidential origo of observation, the speaker. We call this type of firsthand evidential ‘performeroriented evidential’.

The Korean -te- is a firsthand evidential that displays a third type of firsthand evidential. It does not belong to either ‘experiencer-oriented’ or ‘performeroriented’ evidential. The suffix indicates that the speaker has firsthand knowledge about the events/situations. However, it cannot describe situations performed by the speaker himself. It cannot be used to refer to the speaker’s own activity as a volitional agent. It only refers to the speaker’s observation and does not have the speaker’s own event within its referential scope. In other words, the evidential origo of the observation is excluded from the referential scope of the observee. We name this type of firsthand evidential ‘observer-oriented evidential’. Korean is not alone in this type of firsthand evidential. We can find observer-oriented firsthand evidentials in Amdo Tibetan and Qiang as well (Song 2007).13

The three types of firsthand evidentials can be summarized as in Table (2).

In section (4) we discussed subject restriction and first person effect. There we mentioned that firsthand evidentials are preferred in the first person context and when first person occurs with non-firsthand evidentials, it is in the context of lack of awareness, intention or control on the part of the speaker. When we divide firsthand evidentials into three types in table (2), our explanation should change. What has been explained as ‘firsthand evidential’ in the literature is a type of ‘experiencer-oriented evidential’ or ‘performer-oriented evidential’. In these languages, first person situations are usually expressed by firsthand evidentials, and only unawared, inadvertent, or uncontrolled situations of the speaker can be expressed by non-firsthand evidentials.

‘Observer-oriented evidential’ is different from ‘experiencer-oriented evidential’ or ‘performer-oriented evidential’. Observer-oriented evidentials are not used in describing the speaker’s own performance unlike experiencer- or performer-oriented evidentials. As we observed in section (4), the Korean firsthand evidential -te- is not usually employed in first person contexts (examples 26a and 27a). It is also noticed in section (4) that Korean shows a pseudo-‘first person effect’. When the speaker was lacked awareness, intention or control in a situation, -te- can be used in first person contexts (examples 28-30).

Both ‘non-firsthand evidentials’ and ‘observer-oriented firsthand evidentials’ are similar in that they are not allowed to describe first person situations. In similar situations, however, they can occur with first person participants. When the first person participants lack awareness, control or intention of the situation, they are allowed with first person participants. Non-firsthand evidential usage in first person contexts is called ‘first person effect’ in the literature. In this article, we think that ‘observer-oriented firsthand evidential’ usage in first person contexts should be dealt with in the same way. Based on these observations, which were made by dividing firsthand evidential into three subtypes, the definition of ‘first person effect’ can be modified as follows.

The nature of ‘first person effect’ is that evidentials which are inadequate for first person are allowed to occur in first person contexts when the first person participant is not aware of the situation or does not have control of or intention in the situation.

10The ‘evidential origo’ refers to the person from whose perspective a given evidential is evaluated (Garrett 2001:4).  11The forms on the left of the arrow are in an underlying morphophonemic representation and the one on the right is its surface representation (Oswalt 1986:31).  12A reviewer of this article pointed out the possibility that the performative suffix in Kashaya may be a first person marking. According to Oswalt, (1986: 34) the category of person is not developed in the Kashaya verb, but it has a well-developed evidential category. This article follows Oswalt (1986)’s analysis and includes ‘performative’ within the evidential system.  13Comparing Korean, Mongolian and Amdo Tibetan, Song (2007) proposed to divide firsthand evidentials into two subtypes, ‘experience’ and ‘observation’, depending on their information scope. He attached Mongolian to ‘experience’ type and Korean and Amdo Tibetan to ‘observation’ type. Observing the third type, the performative in Kashaya and the personal agency in Central Pomo, this article proposes three subtypes of firsthand evidentials.

In this article we examined grammatical forms of Mongolian and Korean that can be used to describe past situations. The three Mongolian suffixes -laa, -jee and -v are past tense markers, and they are distinguished in their evidential meaning: firsthand past -laa, non-firsthand past -jee and neutral past -v. In Korean, -ess- is a past tense marker, whereas -te- is a firsthand evidential marker, past sensory observation. Though Mongolian non-firsthand past -jee and Korean firsthand -teare different in their evidential property, they exhibit similar subject restriction. They do not usually occur with first person subject. They occur with first person subjects, however, in contexts in which the first person lacks awareness, control or intention of the situation. Since languages with evidential systems prefer firsthand evidential for first person contexts and avoid employing non-firsthand evidentials, the behavior of Mongolian -jee is as expected. Korean -te- is unusual in that it is firsthand evidential, though it is avoided in first person contexts.

To explain such idiosyncrasies of Korean firsthand -te-, this article proposed to divide firsthand evidentials into three subtypes depending on the referential scope of the observee: experiencer-oriented, performer-oriented, and observeroriented. The experiencer-oriented evidential includes first, second and third person within its referential scope, hence there is no subject restriction on it. The performer-oriented evidential is only used to describe the speaker’s own performance. It does not encompass the second or third person in its referential scope and therefore, the second or third person cannot be used as its subject. The observer-oriented evidential excludes the speaker from its referential scope but includes the second and third person. Mongolian firsthand past -laa is a kind of ‘experiencer-oriented evidential, whereas Korean firsthand -te- belongs to ‘observer-oriented evidential’. This article also redefined ‘first person effect’ to deal with similar phenomena observed in ‘non-firsthand evidential’ and ‘observeroriented firsthand evidential’. Evidentials that are inadequate for the first person are allowed to occur in first person contexts when the first person participant is not aware of the situation or does not have control of or intention in the situation.