The semantics of the two causative suffixes -Ci and -Cii in Andong dialect


 Most Korean dialects have only one set of suffixes, -i, -hi, -li, -ki, as causative suffixes where the main vowel is i. On the other hand, Andong dialects of Korean have two sets of causative suffixes; while one set (-Ci) consists of -ˈi, -ˈhi, -ˈli, -ˈki, etc., the other set (-Cii) contains -íi, -híi, -líi, -kíi, etc. This article proposes that the choice between -Ci and -Cii in Andong dialect depends on the degree of causer’s agentivity. -Ci indicates causation characterized by the full agentivity of the causer. On the other hand, -Cii indicates that the degree of causer’s agentivity is relatively low.


Introduction
Gyeongbuk dialects of Korean have more causative suffixes in comparison to other dialects, including Standard Korean. As is well-known, Standard Korean has two groups of causative suffixes; i.e., one group consists of -i, -hi, -li, -ki, and the other group consists of -wu, -kwu, -chwu (Koo et al. 2015: 149).¹ The sentences in (1) are examples of the former group of suffixes -i, -hi, -li, -ki in Standard Korean.
The sentences in (2) are the counterparts of (1) in Andong dialect, which is one of the Gyeongbuk dialects. In Andong dialect, two distinct suffixes, -ˈhy (-Ci) (allomorph of -ˈhi) as in (2a) and -héyey (-Cii) as in (2b), are used. This should be contrasted with the fact that only one identical suffix -hi is used in Standard Korean (1a, b).
(direct causative) stick-ACC lie.down-CI-IMP "Lay {a/the} stick down." (b) ˈemeni-ka ˈcasik-eykey ˈchayk-ul ilk-héyey-nta. (sociative causative) mother-NOM child-DAT book-ACC read-CII-DECL "A mother makes her child read a book." The functions of the suffixes -i/hi/li/ki in Standard Korean are expressed by one group of suffixes like -ˈi/ˈhi/ˈli/ˈki in some cases and by another group of suffixes like -éyey/héyey/léyey/kéyey in others in Andong dialect. In most cases, these two groups of suffixes cannot alternate with each other; for instance, we cannot use -héyey to derive nwup-héyey in (2a) or use -ˈhi to derive ílk-hi in (2b). This fact implies that they have distinct grammatical functions. However, previous studies on causative suffixes in the Gyeongbuk dialects have neither classified the whole allomorphs of those causative suffixes nor accounted for the semantic differences among them. A series of studies on the tones of Gyeongbuk dialects, beginning from Gim (1980), has classified causative suffixes according to their tonal behaviors. However, these studies have not classified the groups diachronically since their interest is restricted to synchronic phenomena. Choi (1993), Kim (2001), and Bak (2004Bak ( , 2007 have argued for diachronic changes of some groups of suffixes, but their arguments are very limited because they do not cover the allomorphy of causative suffixes exhaustively. As for the difference in grammatical functions, Jung (1996) has indicated that -ˈhi and -híi derive two distinct words with different meanings from one verbal stem. He pointed out that both ánc-hi "to get (foodstuff) ready to cook" and anc-híi "to make (sb) sit" are derived from anc "to sit." However, his argument is limited to this particular verb stem and fails to provide a general explanation for other verb stems.
I have attempted to account for the difference in grammatical functions of -Ci and -Cii from both synchronic and diachronic perspectives in two unpublished papers (Ito 2015(Ito , 2016. However, the analysis of these articles had two limitations. First, I analyzed data mostly from diachronic perspectives and did not provide a sufficient synchronic explanation. Second, the sets of data that I used (i.e., The Academy of Korean Studies 1989, Bak 2008) are lists of words without any contextual information although the verb suffixed with a causative morpheme can be used in various situations. This article will scrutinize synchronic differences in functions between the two causative suffixes -Ci and -Cii in concrete situations. New data have been collected in Andong city, Gyeongbuk prefecture, in September 2016 and September 2017. My main consultant was born in 1939, and he has lived in Andong ever since. He graduated from high school and worked as a local public official. This paper is organized as follows. Section 2 claims that apparently numerous surface forms of causative suffixes in Gyeongbuk dialects can diachronically be seen as allomorphs of three morphemes (i.e., -Ci, -Cii, and -Cwuwu). In Section 3, drawing on the classification of causative types by Shibatani and Chung (2002), I will claim that the choice between -Ci and -Cii depends on the situation given. Furthermore, Section 4 proposes that, more concretely, the choice depends on the degree of causer's agentivity. Finally, Section 5 is a brief summary.

Causative suffixes
This section recapitulates the diachronic classification of causative suffixes in Gyeongbuk dialects from my earlier works (Ito 2015(Ito , 2016. Gyeongbuk dialects have numerous surface forms of causative suffixes attached to consonant-ending verb stems (hereafter, simply, consonant stems). These surface forms can diachronically be viewed as allomorphs of three morphemes: one morpheme which has wu as its vowel, i.e., -Cwuwu, and the other two which have i as their vowel, i.e., -Ci and -Cii. -Cwuwu, -Ci, and -Cii are the representative forms of the numerous surface forms for consonant stems. Each of them is the descendant of the reconstructed forms -*Hwú-*Hwu, -*ˈHi, and -*Hí-*Hi, respectively.
Among the six types of change (i)-(vi) above, only "Leveling" is morphological, and all the rest are phonological in nature. Leveling change occurred in the courses of change toward (D), (H) and (I) has much to do with the synchronic phonological process in Gyeongbuk dialects: ye [jʌ] as an underlying form → ey [e] as its surface form, which is exemplified by iki "to win" + -eto "even if" = ikyeto → ikeyto and machi "to end" + -eto "even if" = machyeto → macheyto as described in Lee (2001: 192).
As shown in Figure 2, we may assume that meyk-éyey "eat-CII" in Andong dialect has once been *meyk-íi, which is conjugated as *meykíiko "make (sb) eat and" (in front of a suffix that starts with a consonant, e.g., -ko "and"), *meykíima "if (subject) make (sb) eat then" (in front of a suffix that starts with the vowel u, e.g., -uma "if. then"), *meykéyeyto "even if (subject) make (sb) eat" (in front of a suffix that starts with the vowel a or e, e.g., -a/eto "even if"), and so on.
Since morphological leveling has taken place in analogy with the form meykéyeyto, the paradigm has changed to meykéyeyko, meykéyeyma, meykéyeyto, and so on. As a result, the stem of this verb suffixed with -Cii is reanalyzed as meyk-éyey rather than meyk-íi. This is the mechanism of leveling, which is assumed to have occurred in Andong dialect.
From a diachronic perspective, -Cwuwu, -Ci, and -Cii have developed from the three original forms -*Hwu-*Hwu, -*Hi, and -*Hi-*Hi, respectively. From a synchronic perspective, however, they cannot be treated as morphemes in a strict sense. This is because the surface form of C is not fully phonologically conditioned. Hence, this study treats these "diachronic morphemes" as sets of suffixes that share the same diachronic origin.

Correspondence between causative suffixes and causative situations
In this section, we will, first, overview the three causative situations noted by Shibatani and Chung (2002) and, then, we will examine how the choice from among the two types of causative suffixes whose main   vowel is i is made in a given situation in Andong dialect. Although there is another causative suffix -Cwuwu in Andong dialect, we exclude this suffix from our discussion in the rest of this article because its grammatical function is basically equal to that of -Cwu in Standard Korean.
(3) (a) emeni-ka ai-eykey os-ul ip-hi-ess-ta. According to Shibatani and Chung (2002), three types of causation are characterized as follows. Direct causation (3a) is characterized by an agentive causer, a patient causee, and a spatiotemporal overlap of the causing and caused events. As a result, there is no time lag or spatial multiplicity between two events. Sociative causation (3b) is characterized by an agentive causer, an agentive causee, and a spatiotemporal overlap of the events. Indirect causation (3c) is characterized by an agentive causer, an agentive causee, and no spatiotemporal overlap of the events. Shibatani and Chung (2002) claim that, in Standard Korean, the morphological causative with -i, -hi, -li, -ki, -wu, -kwu, or -chwu only expresses direct or sociative causation as in (3a) and (3b), respectively, and indirect causation is expressed by the analytic (or periphrastic) causative construction with -key ha "COMP do" as in (3c). The data that I have collected also show that indirect causation is expressed only by the analytic form -key ha rather than by one of the causative suffixes in Andong dialect in the same way as Standard Korean. Since we are concerned with the usage of the causative suffixes in Andong dialect, we will limit our discussion to direct and sociative causation in the rest of this article. Shibatani and Chung (2002) have also divided sociative causation into three subcategories: (a) jointaction sociative, (ii) assistive sociative, and (iii) supervision sociative. However, since I was not able to collect adequate data on joint-action sociative causation, only assistive and supervision sociative causation will be discussed in this section.

Basic correspondence
Section 3.2 will examine the basic correspondence between the causative suffixes and the types of causative situation. In the typical direct causative situation, only the single form -Ci is used. In assistive sociative causative situation, either the single form -Ci or the double form -Cii is used. In supervision sociative causative situation, only the double form -Cii is used.

Direct causative
Typical direct causative situations are depicted by a single-form causative suffix in Andong dialect, as shown in (4). The causee is inanimate in (4a), but that in (4b) is animate.
Both examples have agentive causers, i.e., the listener of the imperative sentence in (4a) and salam "person" in (4b), and patient causees, i.e., máktay "stick" in (4a) and ˈnam "other people" in (4b). In either case, the causing and caused events are spatiotemporally overlapped. Thus, the examples in (4a, b) both depict direct causation. In this kind of causative situation, only -Ci is employed, but -Cii is never except in permissive direct causative situations. This exceptional case will be discussed in Section 3.3.1.
In the following are examples of typical direct causatives derived from intransitives by adding -Ci.

Assistive sociative causative
Assistive sociative causatives depict the situation where the causee executes a caused event and the causer assists the execution of that event through a causing event. This type of situation is expressed mainly by -Cii although -Ci is also used on some occasions.

I-NOM
hide-CII-CONJ give-PAST-DECL "(He) asked (me) to let him hide, and (so) I let (him) hide." In (7a), swúm-ki "to hide (v.t.)" is used since the speaker does not care about the causee's volition. In (7b), the speaker explicitly states that the causee has no will to hide; hence, the same form as in (7a) is used. These two cases can be regarded as typical direct causative situations. On the other hand, when the speaker wants to emphasize the causee's strong volition about his hiding as in (7c), the double-form suffix -Cii, i.e., swum-kéyey, is used. When the causee has his own strong volition, all the causer has to do is just to help the causee to hide. Thus, (7c) can be regarded as an example of assistive sociative causative rather than direct causative.


Other examples of causatives used in assistive sociative situations are as follows.

Supervision sociative causative
In supervision sociative causative situations, the causee executes a caused event and the causer simply orders and supervises this execution. In such situations, the double-form suffix -Cii is used without exception.
Other examples of causatives used in supervision sociative situations are as follows.

Nontypical direct causative
As argued in Section 3.2.1, direct causative is marked with -Ci in principle. However, some types of direct causatives are marked with -Cii. These types of direct causatives are not typical from a semantic point of view. In what follows, I will refer to them as nontypical direct causatives. Nontypical direct causatives are subdivided into two types: (i) those in which the causing event occurs voluntarily rather than by the volitional action of the causer and (ii) those in which the caused event occurs in the causee's inner sphere.
 5 The -Cii form anc-héyey is used in either assistive sociative or supervision sociative situations. The meaning in the former is "to help (sb) to sit" and that in latter is "to make (sb) sit (by ordering)."

Permissive direct causative
When the causing event occurs because the causer does not do anything rather than to do something, the double-form suffix -Cii is used in Andong dialect, as shown in (11) (hereafter, I will call nontypical direct causatives of this type permissive direct causatives).

Bodily/mental change direct causative
If the caused event in direct causative situation is an action that occurs inside the causee's body or mind, -Cii is mainly used, as shown in (12) and -Ci is occasionally used.⁷ (hereafter, I will call direct causatives of this type bodily/mental change direct causatives).
(12) (a) áyay-ka kǎmki tul-ess-nunci aphú-ntey child-NOM cold enter-PAST-CONJ ill-CONJ kǒi nwup-héyey-la. dearly lay.down-CII-IMP "(The) child is ill because (he or she) has caught a cold, so lay (him/her) down dearly." (b) Sónca-ka caylóng-ulo ˈhalabeci-lul wus-kéyey-nta. grandson-NOM cute.tricks-INST grandfather-ACC laugh-CII-DECL "{A/The} grandson makes his grandfather laugh with cute tricks." In (12a), the caused event is the event where the causee ayay "child" lies down. The movement of lying him/her down entails a change in the causee's body posture, which is taken for an action executed in the causee's inner sphere. In (12b) as well, the caused event of the grandfather's laughter takes place in his inner sphere; hence, the double-form -Cii is utilized.
The degree of causee's control over the caused event does not affect the selection of the suffix in bodily/ mental change direct causative situations. Thus, as shown in (13), -Cii is used in bodily/mental change direct causative situations, where the causee is unconscious or even inanimate.
 6 Note that -lí in nol-lí-nta is a surface form of -Cii rather than -Ci as the result of (I) in Figure 1 (see footnote 2 above). 7 The condition on the appearance of -Ci is not clear at this point. I will leave this for the future research.
( a) méli kkám-ala. head wash-IMP "Wash (your) hair." (b) méli kkam-kéyey-∅ ˈcw-ela. head wash-CII-CONJ give-IMP "Wash (his/her) hair." The type of construction found in (14b) is called "non-canonical causative construction" by Chung (2006: 110-22, 142-9). She argues that such a construction is not causative in syntactic terms according to the definition by Comrie (1981) since it is not accompanied by an increase in valency. However, she also claims that such a construction can be treated as causative from a semantic perspective because an increase in kinesis is observed in the derivation from (14a) to (14b). In the situation denoted by the verb stem in (14a), the subject's action occurs inside his or her own inner sphere (low kinesis), and in the situation denoted by the derived verb in (14b), the external subject's (or causer's) action reaches the outside of his or her own inner sphere (high kinesis). Washing somebody else's hair requires higher kinesis than washing one's own Change in body posture anc-héyey "to make (sb) sit physically," nwup-héyey ∼ nwúp-hi "to lay down," kkam-kéyey "to close (sb's eyes)" Grooming or body care kkam-kéyey "to wash (sb's hair)," pis-kéyey "to comb (sb's hair)," ssis-kéyey "to wash (sb's body)" Wearing or carrying ip-héyey "to make sb wear," ep-héyey "to put sb on sb's back," pes-kéyey "to take off (sb's clothes)" Ingestion meyk-éyey ∼ méyk-i "to make sb eat," mul-líi "to make sb bite," ppal-léyey "to make sb suck," kwulm-kéyey "to starve," kolh-líi "to starve" Emotive action wus-kíi "to make sb laugh," wul-líi "to make sb cry" Cognition al-líi "to tell" a I have borrowed the names of several types of verbs from Kemmer (1993), but it does not mean that I have adopted his analysis of reflexive and middle voice in this article.
hair. For this reason, a causative suffix is used for (14b).⁸ The present article also treats the "noncanonical causative construction" as a causative construction since the data show that canonical and noncanonical causative constructions have no difference in terms of the selection of suffixes -Ci and -Cii as argued in Section 4.

Pseudo-causative
Here the pseudo-causative construction is defined as a construction that contains a causative suffix morphologically but lucks a decausativized counterpart and has a different meaning from a genuine causative.⁹ In pseudo-causative constructions, the verb stem is attached with -Ci rather than -Cii as shown in (15a).
(15) ( a) kǎy-∅-to twŭ mali méyk-i-nta. dog-ACC-ADD two CL eat-CI-DECL "(I) feed two dogs as well." (b) *kǎy t w ǔ mali-ka mek-núnta. dog two CL-NOM eat-DECL "(int.) Two dogs are fed." In (15a), the suffixed verb méyk-i means "to feed." If this verb is a genuine causative, then the verb stem mek should render the meaning "to be fed." However, the verb mek does not have such a meaning. Of course, the sentence in (15b) is acceptable in the reading: "Two dogs eat," but it is unacceptable under the intended reading: "Two dogs are fed," which is the decausativized counterpart of (15a). Ci-suffixed verbs like méyk-i in (15a) are not derived by causativization in a strict sense. However, the present article also treats the "pseudocausative" as a causative construction since they are considered to be derived by causative morphology from a diachronic perspective although they are not synchronically productive in Andong Korean.
As shown in Table 3, the morphological form of the same verb stem varies depending on the type of causative situation in which it appears. -Ci is used in the pseudo-causative whereas -Cii is used in the nontypical direct causative, which is taken to be a genuine causative in this article.

Causative suffixes for vowel stems
In my earlier works (Ito 2015(Ito , 2016, I claimed that the two causative suffixes for vowel stems -ywu and -ii are parallel to those for consonant stems -Ci and -Cii, respectively. Section 3.5 will examine the validity of this claim. Table 4 shows the correspondence between the suffixes for consonant stems and those for vowel stems. Note that the two suffixes for vowel stems -ywu and -ii are in comparison to the two suffixes for consonant stems -Ci and -Cii, respectively.¹⁰ As shown in Table 4, it can be confirmed that -ii and -ywu correspond to -Cii and -Ci, respectively, in most cases. Nevertheless, in some cases, -Cii is used for consonant stems, -ywu and -wu are used for the  8 I will go back to the question as to why -Cii is selected for noncanonical causatives rather than -Ci in Section 4.3. 9 The causative construction na-nun sasil-ul swum-ki-nta. (I-NOM truth-ACC hide-CI-DECL) "I hide the truth" does not have a decausativized counterpart, i.e., *sasil-i swum-nunta. (truth-NOM hide-DECL) "The secret hides." However, in the definition of this article, the causative swum-ki "to hide (st)" is not taken for a pseudo-causative since its meaning is almost identical to the genuine causative swum-ki "to hide (sb)" as in na-nun chinkwu-lul swm-ki-nta. (I-NOM friend-ACC hide-CI-DECL) "I hide my friend." 10 One other suffix for vowel stems -wu is not discussed here because it is parallel to -Cwuwu, which is not treated in this article. vowel stems rather than -ii. In (16), the suffix -ywu is used for assistive sociative causative where the suffix -ii is expected.

(16)
áyay-tul cwul sé-ywu-nta. child-PL line stand-YWU-DECL "(A teacher) makes pupils line up." This phenomenon shows that the morpheme -ywu has a wider grammatical function in comparison to -Ci. This is presumably because -ywu has an extra morpheme -*Hwu in its reconstructed form -*Hi-*Hwu whereas the reconstructed from of -Ci is simply -*Hi.

Analysis
This section focuses on the core difference in semantic function between -Ci and -Cii in Andong dialect. As noted in Section 3, the choice between the two causative suffixes depends on the situation. In Section 4, I will argue that the choice is made by causer's agentivity. Causer's agentivity is concerned with the degree to which the causer has control over the complex event denoted by the whole sentence. Causer's agentivity is also closely related to the degree of directness with which the causer participates in the whole event. The stronger control the causer has over the whole event, the more directly the causer tends to affect the whole event. In Andong dialect, while the single suffix -Ci is used to indicate the full agentivity of the causer, the double suffix -Cii is used to indicate its partial or relatively low agentivity. anc "to sit" ánc-hi "to prepare to cook" anc-héyey "to make (sb) sit" mek "to eat" méyk-i "to feed" meyk-éyey "to make (sb) eat" nǒl "to play" nól-li "to make fool" nol-léyey "to let (sb) play" pes "to take off (ref.)" pés-ki "to peel" pes-kéyey "to take off" wǔs "to laugh" wús-ki "to be funny" wus-kéyey "to make (sb) laugh" Direct causative (typical) -Ci -ywu a ttú-ywu "to float (v.t.)," sé-ywu "to stand (v.t.)," tháywu "to give a ride" Pseudo-causative -Ci -ywu kh(u)-íwu "to bring up" Nontypical direct causative (bodily/mental change) -Cii (∼-Ci) ii b cwi-éyey "to make (sb) grasp," i-éyey "to put sth on (sb's head)," pi-éyey "to put a pillow under (sb's head)," po-í "to show" -ywu cá-ywu "to put sb to sleep," sé-ywu "to stand (v.t.)" Assistive sociative causative -Cii (∼-Ci) ii nwu-éyey "to make (sb) pee" -ywu sé-ywu "to make (sb) stand"

Causatives marked with -Ci-
In the typical direct causative (Section 3.2.1), -Ci is used. This is because the causer has control over the whole event, including the final result, and, thus, its agentivity is full. In Section 3.4, we have argued that pseudo-causatives can be analyzed with the same scheme as observed in synchronic causative derivations although pseudo-causative does not undergo causative derivation in a strict sense. Suffixes in pseudo-causatives differ from those in genuine causatives, as shown in Table 3. Pseudo-causatives and nontypical direct causatives have different degrees of causer's agentivity. This is natural because their causative meanings are not identical. Pseudo-causations are more like typical direct causatives. The causee of ánc-hi "to prepare to cook"or pés-ki "to peel" is inanimate; thus, the causer has full control over the event. The causee of méyk-i "to feed" or nól-li "to make a fool of"for example, the animal which is fed, or the person who is made a fool ofhas no control over the depicted event. For this reason, the causer has full control and hence -Ci is utilized.¹¹

Causatives marked with -Cii-
Supervision sociative causatives (Section 3.2.3) and permissive direct causatives (Section 3.3.1) are marked with -Cii. In supervision sociative causatives, the causer has only weak agentivity because he or she merely controls the initial stage of the whole event and does not directly execute it but rather supervises the ensuing stages.
Similarly, in permissive direct causatives, the causer's agentivity is weak since he or she does not directly execute any action but rather permits the resultant situation. Shibatani (1978: 310-21) argues that permissive causatives can be divided into positively permissive (積極許容) and passively permissive (消極 許容) causatives. In the former, the causer willfully permits the causee to execute the caused event; in the latter, the causer simply overlooks the execution of the caused event although the causer is responsible for or related to the caused event. Shibatani (ibid.) claims that the causer of positively permissive causatives is agent and that of passively permissive causatives is experiencer. Although the causer of positively permissive causatives is an agent, his or her agentivity is relatively lower than that of factitive (i.e., nonpermissive) causatives, because of the lower directness of execution.

Causatives marked with either -Ci or -Cii
In Section 3.3.2, we have noted that -Cii is used in bodily/mental change direct causatives, in which the caused event occurs inside the causee's inner sphere, i.e., body or mind. In such a case, the causee somehow has control over the caused event.
(17) áyay-ka kǎmki tul-ess-nunci aphú-ntey (=12a) child-NOM cold enter-PAST-ADN ill-CONJ kǒi nwup-héyey-la. dearly lay.down-CII-IMP "(The) child is ill, which might be because (he or she) has caught a cold, so lay (him/her) down dearly."  11 Unlike other causativized verbs derived from action verbs, the pseudo-causative wús-ki "to be funny" is a stative verb. The subject of wús-ki is not agent and is not necessarily animate. Thus, it is not possible to explain the choice of -Ci in terms of the degree of causer's agentivity in this particular case. For example,in (17), the action depicted by nwup-héye "to lay down" can be accomplished only when the causee áyay "child" is cooperative with the causer. If the causee struggles against the causer, then the causer will have difficulty or will fail to lay him/her down. This is how the causee of bodily/mental change direct causatives is different from that of typical direct causatives. The causer of typical direct causatives has full control over the caused event. On the other hand, the causer in bodily/mental change direct causatives has only partial control; the causer affects only the external sphere of the causee because the causer cannot reach the causee's inner sphere. For example, in a situation expressed by the suffixed verb of ingestion meyk-éyey "to make sb eat" in Table 2, a mother can put food into her son's mouth with a spoon (external sphere), but she cannot directly move his tongue and throat muscles, so that he will swallow the food in his mouth (inner sphere). The causing event of bodily/mental change direct causatives can be executed only outside the causee's inner sphere. This is because the causer in bodily/mental change direct causatives cannot have full agentivity because he or she may not control the whole event completely. The causee has some control over the caused event, which is not directly executed by the causer. This is the reason why -Cii is used in bodily/mental direct causatives, as shown in (17).
This explanation is good enough for most bodily/mental change direct causatives, but it is insufficient in some cases where the causee is unconscious or inanimate, and so the causer should be able to control the whole event.
(18) (=10) (a) cam tu-n áki-lul ípucali-lul phyé-ko sleep enter-ADN baby-ACC bedclothes-ACC lay.out-CONJ nwup-héyey-nta. lay.down-CII-DECL "(I) lay out bedclothes and then lay down a sleeping baby." (b) inhyéng-ey ˈos-ul ip-hyée-se nǒli ˈha-nta. doll-DAT clothes-ACC wear-CII-CONJ play do-DECL "(She) plays dressing up (her) doll." It is possible to say that the causee of (18a) áki "baby" unconsciously has some control over its posture even if it is asleep. This control can be considered to be inaccessible to the causer "I." However, it is impossible for the causee of (18b) inhyéng "doll" to have any control over its posture. This state of affairs might be explained by a kind of personification.
In assistive sociative causatives (Section 3.2.2), the causer helps the causee to execute the caused event. The causer's agentivity is not as high as in typical direct causatives since he or she does not have full control over the whole event. At the same time, his or her agentivity is not as low as in supervision sociative causatives because he or she controls the depicted event more directly; while the causer in supervision sociative causatives simply supervises, that in assistive sociative causatives directly helps the causee to execute the action depicted by the verb stem. Thus, the causer of assistive sociative causatives also has a mid-level agentivity as does the causer of bodily/mental change direct causatives.
To summarize, causers in bodily/mental change direct causatives and assistive sociative causatives have a mid-level control over the whole event. The agentivity of these two types of causatives is neither as high as that of typical direct causatives nor as low as that of supervision sociative causatives. Therefore, bodily/mental change direct causatives and assistive sociative causatives are expressed by either -Ci or -Cii.

Causative suffixes for consonant and vowel stems
As argued in Section 3.5, the border of usage between the suffixes for vowel stems -ywu and -ii is not completely identical to that between the suffixes for consonant stems -Ci and -Cii. The set of data found in this study shows that -ywu is more widely used in comparison to -Ci, and, in contrast, -ii is more narrowly used in comparison to -Cii. As shown in Table 4, the suffix -ywu depicts not only typical direct causative and pseudo-causative situations but also bodily/mental change direct causative situations. Whereas the "change in body posture" verb se-"to stand (v.i.)" is suffixed with -ywu, the verbs of "wearing or carrying" (e.g., cwi-"to grasp") and "cognition" (e.g., po-"to see") are suffixed with -ii. Thus, it can be safely concluded that causer's agentivity is higher in "change in body posture" verbs than in "wearing or carrying" and "cognition" verbs. This conclusion seems to be natural because, in events denoted by "change in body posture" verbs, it is simpler and easier for the causer to interfere into the causee's inner sphere. For instance, to move someone's arms or legs in order to make him or her stand up involves smaller degrees of complexity, hence, difficulty than to move someone's body in order to dress him or her.

Typological analysis
As argued above, the "causative" as a synchronically productive grammatical category can be divided into four groups in Andong dialect: (i) typical direct causatives, where the causer's agentivity is full, are realized with the shortest form -Ci; (ii) assistive sociative causatives and bodily/mental change direct causatives, where the causer's agentivity is on an intermediate level, are realized mainly with -Cii and occasionally with -Ci; (iii) supervision sociative causatives and permissive direct causatives, where the causer's agentivity is low, are realized with the longer form -Cii-; (iv) indirect causatives, where the causer's agentivity is very low, are realized with the longest from -key ha. Causative types are more finely divided in Andong dialect than in Standard Korean. Thus, this dialect provides a better example that exhibits the continuum from direct to indirect causatives. Moreover, the system of causative suffixation in Andong dialect corresponds to the typological data. Dixon (2000) argues that the more unmarked a causative is, the more compact form it is encoded with. He also maintains that different degrees of control over the event have much to do with the markedness of causatives: the greater control the causer has, the more unmarked the causative is. In Andong dialect, causatives whose causer has full agentivity (i.e., most unmarked causatives) are denoted with the shortest form, and vice versatypical direct causatives are expressed with -Ci and indirect causatives are with -key ha. Thus, cases in Andong dialect straightforwardly conform to the widely recognized typological tendency.

Conclusion
The Gyeongbuk dialects of Korean have three causative suffixes represented by -Cwuwu, -Ci, and -Cii. In Andong dialect, which is one of the Gyeongbuk dialects, the two causative suffixes -Ci and -Cii indicate different degrees of the causer's agentivity. -Ci is found in typical direct causatives and pseudo-causatives where the causer's agentivity is full. On the other hand, -Cii is used in supervision sociative causatives and permissive direct causatives where the causer's agentivity is low. In assistive sociative causatives, permissive direct causatives, and bodily/mental change direct causatives, -Cii is mainly used but -Ci is also used occasionally. This is because the degree of the causer's agentivity comes in between that of direct causatives and that of supervision sociative causatives.
Interestingly, in Andong dialect, -Cii denotes not only nontypical direct causatives and sociative causatives but also passives. As argued above, the suffix -Cii requires low agentivity of the causer. In this connection, the subject of passives has no control over the passive event because he/she is merely a patient. Thus, it is natural that passives are denoted with -Cii as well as causatives where the subject of the whole sentence has a very low degree of agentivity. This fact implies that there is a continuum from causatives to passives as argued by Keenan and Dryer (1985) and Haspelmath (1990). In Andong dialect, this continuum is segmented into two grammatical subcategories, -Ci and -Cii, depending on the degree of agentivity which the subject of the whole sentence has. The data of Andong dialect shown in this study rather directly exhibit this continuum.