Alternations of Korean

Alternations
Alternation name Description Examples Verbs
C
y This form of the causative comprises about nine different suffixes which appear mainly with intransitive and a few transitive verbs. The fact that this type of causative is built through suffixation has led to the present naming and has been applied in numerous sources such as Yeon (2003). O'Grady (1991) refers to this alternation as 'Lexical Causative', although Sohn (1999: 374) uses this term to refer to inherently causative verbs that are not morphologically derived from retraceable base forms.

The productivity of this alternation seems to be quite reduced in modern Korean, as it can only be applied on a set of about 400-500 Native Korean verbs (O'Grady 1991: 154). Being lexicalised to a great extent, O'Grady (1991: 154) or Yeon (2003: 79) mention that a morphological causative verb form may have a meaning diverging from the verb's base form (see ex. 183, 120 or 155). A couple of morphological causative forms have been given here which have been found in an on-line dictionary, although they had not been known to me prior to this investigation (see ex. 245 or 278). Only transitive and intransitive verbs have morphological causative forms.

As the name suggests, a more or less striking difference between this type of causative and the periphrastic causative is that the latter type is a syntactically analytic construction.

The semantic differences between this type of causative and periphrastic causative constructions have been discussed in numerous publications (see e.g., O'Grady 1991: 154, 172, Sohn 1999: 376 or Yeon 2003: 83ff.). The more 'traditional' conceptualisation is that the causee in a morphological causative construction has little or no agency. Additionally, contrary to the periphrastic causative, this causative implies a direct impact on the thing or person affected by the causer (e.g., in English there is a slight difference between 'John turned the chair around' and 'John made the chair turn around').

As discussed in the comment field of the entry for
jugda, the distinction between what has been labelled as 'distant' and 'contact' causation by Yeon (2003: 83ff.) is by no means that regularly mappable onto the morphological and periphrastic causatives. For example, with verbs such as kkeulhda 'boil' as in ex. 130 the referral to a distance vs. contact-based explanation is rather dissatisfying since the agent (who is typically animate) cannot make water boil in any (physically) 'direct' way.

Additionally, in many cases the periphrastic causative may include the meaning of both 'contact' and 'distant' causation, and in fact for a multiplicity of Korean verbs this is the only option to build a causative construction.

Moreover, what many sources rather tend to neglect is the fact that the construal of an action (and/or its result) may have an impact on the choice of a particular type of causative. It seems that whether one considers the result of an action or the action itself as more relevant to one's concerns - as vague as this intimation may sound to the reader here - can override the actual 'physical' realities of how an action has been carried out, that is, whether some result was caused directly or indirectly. See the comments on
jugida 'KILL' for more.

Causativised intransitive verbs have a NOM-ACC case pattern, and causativised transitive verbs have a NOM-DAT-ACC pattern. Note that ditransitive verbs do not have a morphological causaive form. Some morphological causative forms seem to be used as a compex predicate together with
juda 'give', although prescriptive sources do not acknowledge this fact enough (see ex. 206 or 250).
(3)
Hyeongi janeul mullo chaewotta.
hyeong-i
older_brother-NOM
jan-eul
glass-ACC
mul-ro
water-INSTR
chae-u-eoss-da
get_full-CAUS-PST-DECL
‘My older brother filled the water into his glass.’
19
C
y This causative can be formed by adding the suffix -ge, converbal ending (see Haspelmath 1995) to the verb stem and using the auxiliary hada 'do'. As opposed to what has been termed 'morphological causative' here, the causative meaning is expressed by analytic means. Standing in opposition to existing morphological forms which are not realised by multiple syntactic words, this construction can be identified as being periphrastic (see Brown et al. 2012 for more).

Its applicability shows a rather unclear pattern, since it seems to be almost fully productive, whereas on the other hand emotion verbs as
johda 'be good', museobda 'be scary' do not seem to take part in this alternation, unless these verbs undergo the hada-alternation. Following Yeon (2003:66ff.) on this, in the base form, a Korean emotion verb describes an emotion that is not accessible to external entities, but only to the experiencer, and Yeon concludes that the hada-transitivisation process brings about an 'externalisation' (Yeon 2003: 66) of such emotions.

As mentioned in the entry for the morphological causative, a lot of research has been done on the morphosyntactic and semantic-pragmatic differences of this construction to the periphrastic causative (see entry for the morphological causative for references, or also Lee and Ramsey 2000: 212-215), although research on these differences tend to suffer from a notorious desire to make a clear-cut distinction between these constructions on the one hand, and misconceptions on the nature of periphrastic verb forms on the other.

Especially the Korean periphrastic construction has attracted a great deal of attention regarding the issue of clausality in constructions involving multiple verbs, and various tests have been applied to show that the periphrastic causative involves the embedding of a clause into another that is headed by the auxiliary
hada (see Sohn 1999: 377 or Song 2005 for example, or O'Grady 1991: 188ff. making a finer distinction in biclausal and monoclausal periphrastic causative constructions).

The coding pattern in Korean periphrastic causative constructions has been discussed in Yeon (2003: 87) on the basis of Comrie (1981). Basically, an intransitive, monovalent verb will have NOM-ACC as its causative case frame, and a causativised transitive verb NOM-DAT-ACC. For causativised ditransitive verbs Yeon (2003: 88) claims that the causee is encoded in a postpositional phrase, although for someone who is mainly acquainted with spoken Korean I cannot support this claim. Rather I would say that the additional argument here is marked with accusative case.

Yeon (2003: 90ff.) indeed discusses cases that diverge from Comrie's generalisations, and several researchers (cf. O'Grady 1991: 171) have tried to show that the causee can be marked with nominative, dative or accusative case depending on the volitionality and agency that the causee retains. This is a claim that I can neither refute nor confirm, although I would like to remark here that in meany examples I have given here in this database I have noted that nominative and dative marking on the causee sound at best awkward to me.
(2)
Eomeoniga ttareul ibureul deopkke hasyeotta.
eomeoni-ga
mother-NOM
ttal-eul
daughter-ACC
ibul-eul
blanket-ACC
deop-ge
cover-CONV
ha-si-eoss-da
do-HON-PST-DECL
‘The mother made the daughter cover herself with the blanket.’
90
C
y The benefactive alternation is a periphrasitically realised (see Brown et al. 2012) valency alternation of the applicative type where a verb's valency is augmented by one. It is generally used in situations where somebody does something for the benefit of someone else, or where someone does something on behalf of someone else (but again, for the benefit of this person).

Following Creissels (2010: 30/31) definition, such a 'benefactive applicative periphrasis' is an applicative alternation that licenses beneficiary roles in particular. As Creissels (2010: 35) further explains, a "benefactive NP is licensed by a word that also occurs with a related meaning in constructions in which it clearly has the status of a verb."

Indeed, the benefactive periphrasis in Korean is built through a complex predicate that consists of a converb and the verb
juda 'give' (and its imperative and honorific counterparts which will not be discussed here, see Sohn 1999: 384ff.), which makes it structurally similar to other auxiliary verb constructions in Korean. In terms of the grammaticalisation of this biverbal construction as a monoclausal construction, Sohn (1999: 384) argues that juda with its NOM-DAT-ACC coding frame has lost its ability to independently assign semantic roles, although his evidence cannot be regarded as providing the full picture.

The present dataset suggests that only transitive verbs with the coding frame NOM-ACC can undergo this alternation (but this might rather be due to the very narrow understanding of a syntactic benefactive alternation here, see discussion below), resulting in a NOM-DAT-ACC coding frame in the benefactive, and looking at ex. 306 one might assume that the ditransitive case frame alternates with a double accusative case frame. SING might be the only candidate to show that an intransitive verb can undergo this alternation as well, although syntactically it is not sure whether this verb is actually intransitive or not, due to its structural properties (see entry on SING).

When a complex predicate is built using
juda, three effects can be identified: 1. an overtly dative-marked NP can appear in the clause (COVER, CUT, TIE, POUR, CARRY, SEARCH) 2. an additional NP appears governed by the postposition wihaeseo 'for' or daesin(haeseo) 'on behalf of' (EAT, WASH, RECEIVE among others) 3. there is no syntactic change, and neither a dative-marked NP nor a postpositional NP can appear. (HUG, WASH (bath) someone, SIT DOWN, HELP and practically all intransitive verbs). The last case is true with ditransitive verbs, and occurs with a lot of causative verbs that are not normally used on their own (see SHOW for example). Curiously however, for a decent number of the verbs in the present data set it seems that such a complex predicate with juda is highly preferred in spoken usage as oppposed to the 'simple' form, and as a speaker who grew up with spoken Korean only, I have indicated that some verbs sound plain wrong without the verb juda (see SHOW (visually) or HELP).

As you can see in the list of verbs, only a handful of verbs allows for an additional, dative-marked NP to appear in the clause when the benefactive alternation is applied. Creissels (2010: 49) discusses Shibatani's (2003: 282-3, see Creissels 2012 for references) idea's regarding an apparently very simialr observation made in Japanese, and argues that "in Japanese, intransitive verbs, or transitive verbs whose object NP denotes an object that is not normally transferred to a beneficiary cannot occur in a BAP [benefactive applicative periphrasis] including an overtly expressed beneficiary". Shibatani (2003: 282-3, cited in Creissels 2010: 49) specifies this 'transferral' as a "transfer of possessive control" which is not necessarily the same as the transfer of an object.

Looking at the explanation above, one might understand better why even verbs such as TIE or CARRY can undergo a benefactive alternation in Korean, since objects are not transferred per se here, but rather the possessive control over an object.

However, most studies on benefactives in Korean reduce the description to a dichotomy where there are only complex predicate constructions involving
juda that either allow for an overtly expressed dative NP or not. As mentioned above, an apparently much more frequent possibility is expressing the beneficiary in a postpositional phrase (see ex. 176), or not expressing the beneficiary at all. For the latter case, I have included some examples which have been linked to verbs which have been checked for 'never' regarding this alternation.

Examples such as ex. 349 might be a hint towards a differentiation between a 'semantic' benefactive applicative periphrasis and a 'syntactic' one in terms of the actual correspondence between syntactic exponence and semantic content of an expression. Obviously, I can only 'nudge' to spark some criticism and elucidation here, and would be grateful if the present discussion could lead to research that focuses on this construction alone. The only paper in English that I know of is Shibatani (1994, focusing on semantic valency properties and their comparison between Japanese and Korean), and a much more detailed investigation is desirable.
(101)
Eomeoniga aiege ibureul deopeojusyeotta.
eomeoni-ga
mother-NOM
ai-ege
child-DAT
ibul-eul
blanket-ACC
deop-eo
cover-CONV
ju-si-eoss-da
give-HON-PST-DECL
‘The mother covered the child with a blanket (for it).’
11
U
n This type of coding pattern alternation pertains to a few verbs that have a reciprocal meaning, for example mannada 'MEET' and iyagihada 'TALK'. anda 'HUG' is an ambiguous case where people do not have to carry out an action in a reciprocal manner.

In general, for the verbs above, a NOM-ACC case pattern alternates with a NOM-COM case pattern. The order of NPs from ex. 121 to shows that the comitative-marked argument does not seem to be embedded into the NP of the nominative-marked argument, speaking for a 'true' reciprocal alternation as opposed to ex. 124.

Cases such as ex. 295 (with the verb
nolda 'PLAY') could be seen as reciprocal alternations as well, although here there is no alternation of two coding patterns, but rather the comitative-marked NP seems to be an optional adjunct.
(54)
Jeo eoje sinaeeseo suyeongihago mannasseoyo.
jeo
1SG.HUM
eoje
yesterday
sinae-eseo
city-LOC
suyeongi-hago
SooYoung-COM
manna-ss-eoyo
meet-PST-POL
‘Yesterday I met with Soo-Young in town.’
3
C
y Similarly to morphological causatives, a verb which undergoes this alternation is suffixed with a small set of different, partly phonologically, and partly lexically determined suffixes (see Yeon 2003: 109, for example, or Lee and Ramsey 2000: 206ff.).

The morphology of this type of passive formation is not productive anymore (Lee and Ramsey 2000: 207), and restricted to a few predicates of mostly native Korean origin. Other types of verbs, for example complex
hada predicates simply exchange the verb hada with another verb such as tanghada 'be affected by something'. The later type of 'passive' formation has not been included in this database here (see O'Grady 1991: 48 or Yeon 2003: 108, for example).

As Lee and Ramsey (2000: 208) describe, the rather high degree of lexicalisation of the morphological passive in Korean has sometimes brought along changes in a verb's meaning. These meaning changes have been mentioned in a couple of verb entries, and in the case of HEAR it has been included as a separate entry. Moreover, this lexicalisation has brought along apparent passive forms that do not have any active counterparts anymore (see Yeon 2003: 103).

In a passive clause, the accusative-marked NP of a former transitive verb is marked with nominative and is granted subject status. It is less obvious what happens with the former nominative-marked NP, however. With a passive verb it either seems to be inexpressible or can be expressed with dative-marking (or locative marking on inanimate nouns) on the NP, or through the postposition
euhaeseo 'by' which governs locative case, with the latter possibility sounding rather stilted in spoken usage. Lee and Ramsey (2000: 207) state that the presence of a locative or instrumental-marked NP in a clause forces the agent NP to be expressed in a postpositional phrase.

According to several researchers (Yeon 2003: 109 or Lee and Ramsey 2000: 208) this alternation does not apply to ditransitive verbs, to verbs of sensation, cognition and emotion, verbs which have reciprocal meaning and as a more morphologically motivated feature, verbs whose stems end with
-i. Among verbs with the coding pattern NOM-ACC, HELP for example cannot undergo mopological passive. Researchers that discuss this alternation do not mention the possibility of arguing that the morphological passive applies to verbs which the argument structure <agent, patient> only, where the patient role implies high affectedness. Admittedly though this suggestion does not work for all of the verbs, since NAME for example has a morphological passive form although the 'patient' is not really highly affected.

(257)
Sajeong sajeongeul haedo mari an meokine.
sajeong
explain_implore
sajeong-eul
explain_implore-ACC
hae-do
do-CONV
mal-i
speech-NOM
an-meog-hi-ne
NEG-eat-PASS-MIR
‘Really, no matter how much I explain my reasons to him, no matter I beg him, my words just do not reach him.’
19
U
n This is an alternation that has been mentioned in a range of publications. In a verbal clause, an argument NP in which the head and the dependent are in a possessor-possessum relationship (optionally indicated by genitive marking on the possessor), the possessum receives the case marking according to the function it fulfills within the subcategorisation of the verb (for example, an NP which is the object of a verb will receive accusative marking).

It is not clear whether one can identify a class of verbs which exhibits this type of coding alternation, although all the discussions in the literature revolve around the specific relationship between the possessor and the possessum within the context of a verbal event. Whereas Yeon (2003: 137) remarks that a possessor ascension construction is only possible when both the possessor and possessum are "regarded as affected by the action" (Yeon 2003: 137), O'Grady states that " 'Possesso ascension' alternations are found in theme NPs naming a whole-part relationship" (O'Grady 1991: 68; although O'Grady himself shows examples where the possessor ascension construction involves NPs that do not seem to be themes, see below), and therefore it remains to be elucidated what could be the common point between all the verbs to which this alternation pertains.

Curiously, there is another possibility than the one mentioned in the first paragraph. With an intransitive verb, both constituents of an original possessor-possessum NP in subject function can be marked with nominative case (ex. 316, 184), or with a transitive verb, both constituents of an original possessor-possessum NP in object function can be marked with accusative case (ex. 253). Interestingly, for a transitive verb, this way of alternative encoding seems to possible only for the object NP, and not for the subject NP. Moreover, this alternative way of marking for object NPs does not seem to be possible for accusative NPs of verbs with NOM-DAT-ACC pattern. Also, intransitive verbs which have a NOM-NOM case pattern (which in turn is said to alternate with DAT-NOM, see entry for
jota 'LIKE' and museobda 'FEAR') do not seem to allow for this alternation where one would have a coding pattern with three overtly nominative-marked NPs.

O'Grady (1991: 72) gives a highly unnatural and questionable example (ex. 315) to argue that in a possessor ascension construction the bearer of the theme/patient role changes from the former possessum to the possessor, and accordingly (hence the term 'ascension'), only the possessor in a possessor ascension construction can become the subject of a corresponding passive construction. Note therefore the (from a Indoeuropean perspective) rather strange coding pattern with NOM-postpos_adjunct-ACC V, which I would reject categorically, although being discussed by several scholars regardless of whether these constructions are actually poduced in speech (cf. Yeon 2003: 179).

A further observation in this construction is that in a possessor ascension construction as in ex. 170 the possessum NP cannot take on a modifier to its left (for example,
dukkeoun thick.ATTR) which would actually be possible in a 'non-ascended' construction. O'Grady (1991: 75) mentions that accordingly, the possessum NP underlies restrictions in terms of its referential specificity here, and more curiously, he discusses the possibility (following Choi 1988: 62, see O'Grady 74 for references) that the second accusative-marked NP in a possessor ascension construction undergoes a categorical change to an adverbial NP, which therefore cannot take on adnominal modifiers as dukkeoun above.

This proposal surely is interesting given the fact that the 'adverbial' status (alhough using this term in a more semantically oriented way) of an accusative-marked NP has been discussed elsewhere in this database (see way/number alternation), although as discussed in the next paragraph below, I am not sure how this adverbial status of an NP and its accusative-marking could be considered in conjunction with the fact that in passive clauses it changes to nominative case. Whatever the case though, a more obvious fact has not been mentioned in the literature, namely the observation that in a possessor ascension construction, the two NPs with identical marking may not be scrambled around, but the original possessor-possessum order has to be maintained. In this light, it is interesting that this Korean does not seem to have been discussed within the possibility of exhibiting Suffixaufnahme (Plank 1995), and maybe this is a way to look a this construction.

As opposed to examples such as ex. 315 I would intuitively say that in a passive clause the two accusative-marked arguments in a possessor ascension construction change to nominative-marked arguments, similarly to the 'recipient double accusative alternation' described here (ex. 287). Arguably this raises the question of how syntactic theories could capture this alternation, and O'Grady (1991: 134ff.) proposes as regarding the former (optionally genitive-marked) possessor NP as sort of a focus NP that has been added at the left edge of a simple passive clause of which the possessum is the subject, a view that I can neither reject nor support here due to a lack of sufficient knowledge of the the theory proposed by O'Grady (1991).
(169)
Meonjeo sagwareul kkeopjjireul kkanda.
meonjeo
first
sagwa-reul
apple-ACC
kkeobjil-eul
skin-ACC
kka-n-da
peel-PRS-DECL
‘First, one peels the skin off the apple.’
26
U
n The existence of a double accusative coding pattern in Korean is well-known (Sohn 1999: 280, Yeon 2003: 136), and various constructions alternate with the present, double accusative coding pattern. The accusative-marked NP that is not accusative-marked in the alternative pattern seems to have a different status from one double accusative construction to another which is why they are treated differently here.

This alternation has been reported for ditransitive verbs with the coding pattern NOM-DAT-ACC including benefactive complex predicates with
juda, where the recipient/beneficiary
argument can be alternatively marked with an accusative case marker. Whereas in the NOM-DAT-ACC coding pattern the dative argument could follow the accusative argument - for example, whenever the dative-marked argument stands in preverbal focus position and not the accusative-marked argument - in the NOM-ACC-ACC coding pattern the word order seems to become stricter, with the accusative-marked recipient NP preceding the accusative-marked theme argument. It has been reported that in elicitation sessions some consultatns may reject this coding pattern as ungrammatical, and some other may acknowledge their existence, while at the same time remarking that it somehow sounds awkward (see O'Grady 1991: 53, for example, where this alternation is called 'dative advancement'). I assume that most of the time, this must be due to the lack of context in single-sentence elicitation and grammaticality checks, since I myself find this construction awkward sometimes although I am sure that I am using it frequently. Curiously enough - given the fact that having overt case marking (and overt NPs!) in every sentence in Korean seems unnatural - it has not been studied whether these mixed results from elicitation could be due to the fact that Korean case markers may bear some information-structural functions.

The interaction with Korean case marking, case marking alternations and information structure has not been researched extensively, and it could be very interesting to look at whether the positional or prosodic information structure functions such as specific focus intonations or focus positions are still applicable in alternative coding patterns such as this double accusative construction.

When a verb with a double accusative coding pattern is passivised, then this may result in a double-nominative coding pattern with the corresponding passive verb, as in ex. 314. Again this double-accusative to double-nominative change seems to pose some problems fo syntactic theory, and elicitation with consultants seems to lead to varying results. Please refer to O'Grady (1991: 134ff.) for more.

Note that BEAT is an interesting case where a NOM-INSTR-ACC alternates with a double accusative coding pattern, but interestingly, only an extremely restricted set of lexemes seems to appear in such a construction, see ex. 254.
(241)
Neo yaereul wae don dallago hal ttaemada doneul jwo, wae?
neo
2SG
i-ae-reul
this-child-ACC
wae
why
don
money
dal-rago
give.IMP-QUOT
ha-l
do-ATTR
ttae-mada
time-every
don-eul
money-ACC
ju-eo
give-PLAIN
wae
why
‘Why the hell do you give him money every time he asks for it? Why?’
3
C
y Similarly to the morphological passive, the periphrastic ji- passive reduces the valency of a verb by one, although it seems to be applicable to a wider range of verbs, sometimes making it the only available option to build passive forms. For some inransitive verbs such as ttwida this periphastic form seems to exist as well, creating some sort of impersonal predicate with an agent implied, but not overtly expressed. This kind of phenomenon does not seem to be well-studied in Korean, however.

As to the diachrony of this construction, Yeon (2003: 111) mentions that "Historically, the verb
ci- came from the verb, ti-, which means 'to fall' (Bae 1988: 112 [SK: see Yeon 2003 for this reference]). It is generally accepted that the passive auxiliary verb ci- is historically same as the main verb ci- which has various meanings, but the prototypical meaning common to the various usages is a change of state (see Lee 1993: pp109-110 [SK: see Yeon 2003 for this reference])." Note that this auxiliary verb has many other functions in constructions where the valency of a verb is not altered (see Yeon 2003: 112/113 for a litle overview).

Researchers such as Yeon (2003: 119ff.) have long been trying o figure out the semantic differences between the morphological passive and the periphrastic passive, and different factors seems to play into each other. The most unproblematic case for illustration might be where a verb has both the possibility to undergo morphological and periphrastic passive alternation.

Ex. 158 shows the morphological passive form of
nohda, where the event of being put somewhere seems to have happened spontaneously (at least seen from the speaker's perspective), and more importantly, it is assumed that the agent is unknown, and the existence of that agent is not even presupposed. As a conseuquence, and agent cannot be expressed here.

Ex. 157 by contrast shows a sentence with a periphrastic passive, where it is implied that a feeling cannot be put away (translated here with an active sentence) even if one wanted to, and therefore one can conclude that the periphrastic passive tends to retain some volitionality and agency with regard to the instantiation of an event. As opposed to a sentence with a morphological passive, some unknown agent seems to be implied in a periphrastic passive construction (see cf. 94, 233), and this is how some researchers such as O'Grady (1991: 49) call such a verb a 'middle verb' (although this term has been applied to a morphological passive form), and this degree of retained agency seems to correlate with retained control of the agent over the patient.

Accordingly, Yeon (2003: 120ff.) states that in Korean there are verbs which describe weather, nature or perception events (see HEAR or SEE in this database) that undergo morphological passivisation, whereas these verbs cannot be inflected for a periphrastic passive, which Yeon takes as evidence to say that periphrastic passives can only be applied when some control over the patient can be maintained.

As you can see, many examples for this type of passive have been translated into English with an active, modal counterpart, and this is maybe how I tried to express the volitionality that is implied in most of the periphrastic passive sentences. Interestingly, other Koreanic languages such as Jeju for example have developed constructions similar to this periphrastic passive in Korean where the auxiliary
ji is used to express deontic modality: al-a ji-keu-nya know-CONV AUX-IRR-Q means 'You think you can understand this?' in Jeju (see Kang 2007 for a little overview).

Unfortunately, literature on Korean tends to generalise too much on the differences between the morphological and periphrastic passive, and it could well be that in spoken language both versions are used interchangeably, with the periphrasticpassive slowly supplanting morphological forms. As an example, ex. 247 shows the periphrastic passive form of
jjijda 'TEAR', and it is little convincing here to argue for volitionality and agency in terms of the event of shoes getting torn.

Still, this might be due to the fact that arguably, when a verb can employ only one passive variant, then it is probable that one and the same construction can exhibit the semantic differences from above depending on the context (see ex. 218 where the periphrastic passive does not imply an agent). In fact, for the verb TEAR I was not really aware of the fact that there was a morphological passive form of TEAR, and this might have fostered my usage of a periphrastic passive form which includes that for agentless passives. Therefore, frequency in usage should be considered when looking at the differences between these two variants, and a corpus-based approach is still pending.

As another argument that should be seen as a warning from being too over-generalising is the fact that for some verbs such as
deopda 'COVER' it is indeed quite difficult to argue for a semantic difference between the two morphosyntactically different types of passive.

Yeon (2003: 123) suggests that the expressibility of an overt agent phrase in a sentence depends on the verbal semantics of a passive verb. Agents in some passive sentences in Korean can either be marked with dative (animate nouns), locative (inanimate nouns) or a postpositional phrase (see entry for morphological passive for more). He distinguishes between resultative and processive passives, where the former depicts a situation where the result of an action is salient (allowing for no agent or only a 'natural force', inanimate agen), whereas the latter a situation where the process (albeit being a vague notion, see ex. 152) of an event is salient.
(286)
Eoje neomeojyeotteoni mureubi da kkajyeotta.
eoje
yesterday
neomeoji-eoss-deoni
fall_over-PST-DECL
mureup-i
knee-NOM
da
all
kka
peel.CONV
ji-eoss-da
PASS.AUX-PST-DECL
‘I grazed my knee yesterday when I fell over.’
27
C
y This alternation seems to apply only to verbs of emotion and verbs of tactile sensation, which have in common a NOM-NOM case pattern that has often been claimed to alternate with a DAT-NOM case pattern, a pattern that I cannot accept as entirely acceptable, at least in informal, every-day usage of Korean.

The emotion/sensation verb in its converbal form can be combined with
hada to a complex predicate (see comment in ex. 308 which curiously changes the case pattern of the whole complex to NOM-ACC.

Several sources (see Yeon 2003:66ff. and Sohn 1999: 382) have mentioned contraints in Korean where as a part of a speaker's socio-cultural, meta-linguistic knowledge, a speaker is considered to be unable to "have direct access to a third person's subjective internal feelings" (Yeon 2003: 66) which is why emotion (and tactile (?) sensation) verbs such as the ones given here can usually only be used with first person (see Evans 2010: 74 for a similar account) or questions addressed to second person when used on ther own and not in reportative constructions.

As Yeon (2003: 65) shows, a Korean verb of emotion or sensation cannot be used in imperatives or volitional future in its base form, whereas with verbs that have undergone this alternation it becomes possible (see ex. 309, for example). Especially the latter evidence suggests that this type of alternation results in the higher volitionality of the experiencer as observed frequently in distinctions between verbs such as the English SEE and LOOK. As a consequence, Yeon concludes that the hada-transitivisation process brings about an 'externalisation' (Yeon 2003: 66) of such emotions, although this kind of explanation unfortunately ends up sounding quite vague, and a joint psycholinguistic, morphosyntactic analysis is required to elucidate this aspect.

Whether the accusative-marked argument can be regarded the grammatical 'object' of a hada-alternated verb is not clear, although as a tentative observation, passivisation cannot be applied on this construction.
(68)
Nunaneun keun jimseungeul museowohanda.
nuna-neun
older_sister-TOP
keu-n
big-ATTR
jimseung-eul
animal-ACC
museob-eo=ha-n-da
be_afraid-CONV=do-PRS-DECL
‘My older sister fears big animals.’
4
U
n Some intransitive verbs show an alternation between a NOM-INSTR and a LOC-NOM coding pattern. In the first variant, the nominative-marked NP represents the location, and the instrumental-marked NP the instrument (in a broad sense, see FILL, ex. 175). In the second variant, the location is marked with locative and the instrument with nominative case (ex. 227). The word order seems to play a role in the LOC-NOM alignment, see ex. 231.

With transitive verbs it is the object status of the arguments that alternates between the patterns. As for intransitive verbs, interviews with consultants point towards a greater 'unmarkedness' of a NOM-LOC-ACC case pattern which is why this one has been chosen as the basic one, albeit being a somewhat arbitrary decision. Here, the instrument is encoded with accusative case, whereas the location is marked with locative case (ex. 230). This pattern alternates with NOM-ACC-INSTR where the location is marked with accusative case, and the instrument with instrumental case (ex. 1). Apparently, it is mainly the coding pattern where the instrument is marked with accusative case (ex. 230) that one can apply the benefactive alternation (ex. 101) on, and applying the benefactive alternation on an NOM-ACC-INSTR resulting in a NOM-DAT-ACC case pattern with the locational NP in accusative case, sounds slightly awkward. However this might be due to my personal usage preferences instead of some underlying linguistic restrictions.

Several researchers (e.g., Yeon 2003: 196) have mentioned that such a case alternation brings about a holistic/partial-affectedness interpretation, although this is questionable.

Passive alternation can be applied on both coding patterns, which shows that the accusative-marking on an argument is indeed a sign for its object status.
(1)
Eomeoniga banjugeul kichintaweollo deopeusyeotta.
eomeoni-ga
mother-NOM
banjug-eul
dough-ACC
kichintaweol-ro
kitchen_towel-INSTR
deop-eusi-eoss-da
cover-HON-PST-DECL
‘Mother covered the dough with a kitchen towel.’
3