Introducing arguments beyond the thematic domain

. Extensive research has focused on how external and applied argument introducing heads such as Voice (Kratzer 1996), Appl (Pylkk¨anen 2008), and i * (Wood & Marantz 2017), an overarching term for Voice and Appl, are represented inside the thematic domain (below TP). A question arises as to whether Voice or Appl can re-side outside the thematic domain (above TP). This work provides empirical evidence from Korean in suggesting that an argument can be introduced by Voice/Appl ( i *) in the CP periphery. Specifically, it lends support to the claim that the discourse participant ADDRESSEE is represented in syntax (Haegeman & Hill 2013; Miyagawa 2017, 2022; Portner et al. 2019 among others). In this regard, this work draws parallels between the thematic domain and the speech act domain, which have been considered to be two separate domains.

Here, I argue that the specifier of an i* is privileged for HON case-assignment. The absence of *HON.ACC on DOs follows accordingly: a DO is not introduced by Voice/Appl (i*) in its specifier position. In other words, a DO is not an external or applied argument. The current analysis also provides an account for the presence of HON.VOC on the addressee: the addressee is realized in the specifier of an i* above TP. Under this approach, the head that hosts the addressee in the CP domain, call it XP ( ...

X (i*)
A central claim of this work is that the addressee is a part of syntax, just like subjects and IOs, which is eligible for honorific case assignment. More crucially, I highlight that an argument can be introduced outside the thematic domain (above TP).
This work is organized as follows. Section 2 focuses on the basic distribution of Korean case markers and how various types of arguments are case-assigned. Section 3 puts forward the claim that case can be assigned to an argument introduced outside the thematic domain. Here, I show that parallels can be drawn between the thematic domain and the speech act domain. Section 4 concludes.
2. Case assignment inside the thematic domain.
2.1. T ASSIGNS NOM AND VOICE ASSIGNS HON.NOM. NOM and HON.NOM are often associated with the subject of a clause. The empirical picture becomes more complicated, however, when psych verb constructions and case stacking are given consideration. In this section, I argue that NOM and HON.NOM originate from different syntactic heads. Specifically, I posit that NOM is assigned from T and HON.NOM is assigned from Voice following Lee & Nie (2022).
Both NOM and HON.NOM appear on the subject of various constructions, including unaccusatives, passives, unergatives, and transitives. This is demonstrated in (4) (7). In order for (7a) to preserve its topic interpretation, NOM has to be absent. In (7b), however, HON.NOM is present.
(topicalization) (8) shows that NOM is obligatorily realized with HON.NOM in the presence of the negated copula anila inducing contrastive focus (see also Lee & Nie 2022;Schütze 2001 Switching the order of HON.NOM and NOM on halmeni 'grandmother' is not possible in (8). As shown in (9), HON.NOM (-kkeyse) must precede NOM (-i or -ka, depending on the phonological context). To put it in another way, HON.NOM is always the inner marker and NOM is always the outer marker. The co-occurrence of the same marker is also ruled out.
(9) a. Halmeni -kkeyse -ka b. * Halmeni -ka -kkeyse c. * Halmeni -kkeyse -kkeyse d. * Halmeni -ka -ka As we have seen in (5)-(8), the distributions of NOM and HON.NOM are not identical. I argue that NOM and HON.NOM differ with respect to where they appear in the syntax. First, I adopt the standard assumption that T assigns NOM. A non-honorified subject undergoes movement to Spec,TP and receives NOM. This is schematized in (10). In (9), we saw that HON.NOM-NOM is possible whereas *NOM-HON.NOM is not. Instead of T, I argue that HON.NOM is assigned by the external argument introducing head Voice (Kratzer 1996). In fact, this is one of the main claims put forward in this work. I posit that honorified subjects in unaccusatives and passives undergo movement to Spec,VoiceP in order to receive HON.NOM from Voice. Presumably, [+HON] on honorified subjects is checked in this position. This type of approach is consistent with Legate's (2003) analysis in that the edge of VoiceP can be a derived position. 3 While HON.NOM appears to share some typical commonalities with inherent case, there is precedence in the literature suggesting that inherent case can at times behave like structural case. Kayne (2004), for instance, argues that movement is involved in assigning what appears to be an inherent case in French transitive causative constructions, namely the prepositional dative caseà. Under Kayne's approach, the prepositional dative case is assigned to a causee via Internal Merge (IM) rather than External Merge (EM). This, in many ways, shares derivational commonalities with how HON.NOM in Korean is assigned in unaccusative and passive constructions. (12) fleshes out the details: (2019) provide independent evidence that Voice can host an honorific feature. Their evidence is obtained from verbal root suppletion. Chung (2009) initially observes that the verb kyey∼eps∼iss 'to exist' in Korean is a suppletive triplet, which is sensitive to the presence of negation (NEG) and HON: (13) Suppletive triplet (kyey∼eps∼iss 'to exist') a.
According to Choi & Harley, the root suppletion rules in (13) are modulated by structural locality. Adopting their analysis, I argue that the suppletion of the verb 'to exist' triggered by NEG is blocked by HON. 4 This is because HON is structurally closer to the verbal root than NEG is. (14a) and (15a) show that eps is illicit in the presence of HON which is realized as -si. (14b) and (15b) show that kyey has to be realized instead. Taking the blocking effect in (14a) and (15a)  give.HON-PST-DECL Yuli gave grandmother a book. (ditransitive) Based on the current assumption that only external and applied arguments are eligible for HON-sensitive case assignment, it is predicted that causees in transitive causatives should receive either DAT or HON.DAT. This prediction is borne out as shown in (18) (2008), ApplP is realized either above or below vP. Adopting Kim (2011a,b), I assume that there are at least two types of ApplPs in Korean. They are High and Low ApplPs. I posit that Appl can assign HON.DAT or DAT. Both honorified and non-honorified IOs are eligible for dative-assignment in Spec,Low ApplP as shown in (22). Marantz (p.c.) adds weight to this claim by discussing the empirical facts in Georgian causatives. Under this view, intransitives embedded under causative and simple transitives receive a parallel derivation in syntax. Specifically, the causee of intransitives and the DO of simple transitives are both realized as the complement of the verb. Following this line of research, I posit that the causee is externally merged as the complement of an intransitive predicate in Korean causatives (see also Den Dikken 2006 for an argument that Italian causees are realized in a complement position). 6 This accounts for the lack of dative assignment for causees under unergatives and unaccusatives. That is, DAT and HON.DAT can only be assigned by Appl in a Spec-head configuration. The assignment of ACC is possible if we adopt Burzio's generalization (Burzio 1986). The distribution of ACC in Korean will be fleshed out in section 2.3. Kim (2011a) assumes that there is an ApplP which is higher than High ApplP. This is referred to as Peripheral ApplP. Peripheral Appl introduces an experiencer subject. We take this Appl as a (HON.)DAT-assigner as shown in (25). Based on the empirical data provided in this section, the generalization that emerges for datives seems to be quite straightforward. As long as an argument is realized in Spec,ApplP, the argument is DAT or HON.DAT-marked. give-PST-DECL Kim gave Yuli an apple.

ACC BUT
(ditransitive) As Kim & Chung (2015) and Lee & Nie (2022) have pointed out, there is no *HON.ACC even when a predicate semantically associates the DO with honorificity or deference. (27)  While DOs can in principle be honorified, the lack of *HON.ACC on DOs or any other constituents in the grammar for that matter suggests that the distribution of HON-sensitive case markers is restricted. The basic distribution of plain ACC associated with a DO follows Burzio's generalization (Burzio 1986). This accounts for the realization of ACC in (26) and (27). (28) Burzio's generalization (Burzio 1986: 178) All and only the verbs that can assign a theta-role to the subject can assign accusative Case to an object.
The lack of *HON.ACC can be accounted for if the assignment of HON-sensitive case is limited to a Spec-head relation between an argument and an i* (Wood & Marantz 2017). The current analysis is also compatible with Marantz (2022) where v is viewed as an i*. So long as an i* establishes a Spec-head configuration with an argument, the analysis remains the same: (29) An honorified argument receives HON-sensitive case from i* iff the argument and i* form a Spec-head configuration.
If (29) is on the right track, the absence of *HON.ACC follows quite straightforwardly. That is, the Head-complement relation established between a DO and v does not satisfy the necessary condition that triggers the realization of an HON-sensitive case marker. Here, it is worth noting that Korean allows multiple ACC constructions (MACs). I address how MACs are derived based on (29). Jung & Miyagawa (2004) argue that both IO and DO can be ACC-marked resulting in MACs. However, both instances of ACC are lost when passivization takes place. This is demonstrated in (30) give-PASS-PST-DECL John was given a book. (Jung & Miyagawa 2004) Based on the empirical picture provided in (30), Jung & Miyagawa (2004) claim that ACC originates from v. When v is intransitive or is associated with a non-causative construction, it loses the ability to assign ACC completely. This is in many ways consistent with Burzio's generalization.
(31) Tree for (30a) based on Jung & Miyagawa (2004) vP The absence of *HON.ACC is due to the lack of a Spec-head configuration between an argument and an i* that assigns its case (in this case Voice). I update the structure given in (31) and put forward (32) as the configuration reflecting MACs under a VoiceP/ApplP analysis. In (32), Voice assigns ACC downwards to IO and DO respectively. 8 (32) Tree for (30a) adopting VoiceP (Kratzer 1996) and ApplP (Pylkkänen 2008) (32) is in line with the overall analysis of this paper: Voice assigns ACC downwards and thus neither IO nor DO forms a Spec-head configuration with Voice. Hence, the absence of *HON.ACC is accounted for.
3. Case assignment outside the thematic domain. The current proposal makes predictions about where arguments are introduced in the syntax. Specifically it predicts that arguments can be introduced outside the thematic domain. The Korean vocative marker (VOC) has alternating forms (ya∼ø) depending on the honorificity of the addressee. This is shown in (33). Honorification also correlates with the presence of the clausal politeness marker -yo. In the presence of plain VOC (-ya), the clausal politeness marker -yo cannot surface as shown in (33a). In the presence of HON.VOC (ø), however, -yo surfaces as in (33b). go-PST-DECL-YO Grandmother, Sarah went home. Choi (2016) and Yim (2016) argue that the clausal politeness marker -yo is spelled out in the head SA of SAP (Haegeman & Hill 2013) if the addressee is honorified. Just like how Voice and Appl introduce thematic participants in their specifier position, then, SA introduces a speech act participant in its specifier position. Following this line of research, I argue that ya (VOC) and its honorific counterpart ø (HON.VOC) are associated with the same head. For now, I remain agnostic about the label of the head itself as it has received different names in the recent years: SA of SAP under Haegeman & Hill (2013), c of cP under Portner et al. (2019), Addr of AddrP under Miyagawa (2022). Regardless of what the label is, I argue that it is an external/applied argument introducing head (i*) that assigns VOC and HON.VOC to the addressee similar to the i* that assigns DAT  Under this view, the addressee is introduced as an applied argument in the syntax. This suggests that there is a type of Appl in the left periphery which is distinct from Pylkkänen's high and low Appls in the thematic domain (see also Tsai 2018). According to Speas & Tenny (2003), the addressee receives a p(ragmatic)-role which is similar to a theta/semantic-role (see also Akkuş & Hill 2021;Chomsky 2021;Haddican & Etxeberria 2022). This opens the possibility that the addressee can be assigned a semantic-role and licensed case like any applied argument in syntax. Cross-linguistic evidence that an addressee is introduced by i* comes from Meadow Mari (Uralic). Burukina (2020Burukina ( , 2022 provides empirical data suggesting that an argument can be externally merged in embedded Spec,CP. Burukina (2020Burukina ( , 2022 points out that this argument is an addressee embedded under a speech-act verb such as kalas 'to tell'. In Burukina's term, the argument (an addressee) bears the 'semantic properties of an obligation holder and a goal of communication'. It is also described as the 'intermediary that receives the original message'. Under her analysis, the argument is assigned DAT from C. An example is provided in (36). According to Burukina,(36) is referred to as a double dative construction. kalas-en. tell-PST2 Maša told us for Petja to come. (Burukina 2022) (37) shows that double datives are prohibited in the matrix clause. This suggests that an additional DAT-marked argument, if there is one, belongs in the embedded clause.
(37) M@j I Maša-lan Maša-DAT (*t@-lan-@t) you-DAT-POSS.2SG vurgem-@m clothes-ACC nal-@n-am. buy-PST2-1SG I bought Maša clothes. / I bought clothes for Maša, on her behalf. (Burukina 2022) Double datives are also prohibited in embedded finite subjunctive clauses as shown in (38). Instead, they are only possible when the embedded clause is an infinitival as in (36). Note that idiomatic interpretations are preserved under raising constructions, but not in control constructions. Hence, (39) suggests that double dative constructions are control constructions. This implies that PRO is the subject of the embedded clause. Burukina (2020Burukina ( , 2022 concludes from her findings that the second dative DP (the addressee) is base-generated in Spec,CP and that it is the closest DP that c-commands PRO. Hence, the second dative DP is the controller of PRO.