Korean adopts a case system that includes nominative (nom), dative (dat), accusative (acc), and vocative (voc). Most of these case markers have an honorific counterpart (i.e., h.nom, h.dat, and h.voc), as shown in Table 1. acc, however, does not join this alliance as there is no reserved morphology for h.acc. The absence of h.acc raises a theoretical question: is this gap a matter of syncretism in morphology, or does it reflect a deeper constraint imposed by syntax?

We will see that the missing h.acc is not an accident. While morphological syncretism, at a first approximation, seems to be on the right track, we will see where this assumption faces empirical challenges by examining the distribution of the case markers presented in Table 1.1 Moving forward, I propose that honorific (h) case markers are sensitive to phases. Adopting Chomsky (2000; 2001), McGinnis (2000), and Ershova (2024), I assume that vP, ApplP, and CP are phases. I argue that an honorific head (H) selects a phase as its complement and H agrees with its closest c-commanding [+human] nominal argument. In cases where agreement is established between H and a [+human, +h(onorific)] argument, [+h] is overtly realized. In cases where H and a [+human, –h(onorific)] argument undergo agreement, [–h] is not overtly manifested and remains phonologically null.

This work is organized as follows. Section 2 lays out the relevant empirical facts and addresses the limits of a syncretism-based analysis. Section 3 develops a syntax-driven account of the phenomenon under investigation. Section 4 demonstrates that the proposed analysis has implications for other aspects of the grammar. Section 5 concludes.

At first glance, morphological syncretism appears to be an obvious choice of analysis. Based on this view, it is more or less an accident that acc and h.acc converge in their phonological form. Both acc and h.acc surface as -(l)ul due to syncretism. This assumption, however, is at odds with the empirical fact shown in (1). Similar to Icelandic, Korean has psych verb constructions where the subject is dat-marked and the direct object (DO) is nom-marked. Here, nom on the DO cannot be replaced with h.nom, suggesting that the DO cannot receive an h case marker in the first place.

The same restriction holds in multiple nominative constructions (MNCs), as demonstrated in (2). The DO can be marked nom, but it cannot be marked h.nom.

Negated copula constructions display the same pattern. nom on the DO cannot be replaced with h.nom.

The empirical evidence provided above undermines a syncretism-based analysis. Even in the absence of syncretism, h.nom fails to surface together with the DO. Morphological syncretism, in this respect, does not explain the broader empirical picture encompassing (1)–(3). In contrast to the DO, the subject can be associated with h.nom, as demonstrated in (2), (3), and (4). An indirect object (IO) can also be accompanied by an h case marker, namely h.dat, as shown in (4). The DO, as we have discussed, is the odd man out.

The overall pattern stays the same even in simple transitives, where the DO does not receive an h case marker. (5) nicely fits into the general restriction observed in (1)–(4).

At this point, a bigger question needs to be addressed: why are h case markers disallowed specifically with the DO? In other words, what is special about DOs that prevents them from receiving an h case marker? Rather than treating the absence of h.acc as a mere gap in the case paradigm, I argue that the puzzle calls for a deeper explanation, which follows from an independently motivated syntactic restriction. Specifically, I claim that the syntactic position occupied by a DO is ineligible for any kind of h case licensing, which captures the facts presented in (1)–(5).

Korean exhibits an asymmetry: subjects and IOs can bear an h case marker, but DOs cannot. The lack of h case marker on DOs follows from an independently attested concept in syntax. A DO fails to receive an h case marker because it cannot occupy a structural position eligible for H agreement due to a phase boundary.

Adopting Chomsky (2000; 2001), McGinnis (2000), and Ershova (2024), I posit that vP, ApplP, and CP are phases. I argue that every phase is selected by H. Agreement is successfully established when the subject or the IO remains in the same phase domain with H, as shown in (6) and (7).2 Based on my analysis, v and Appl are phase heads and their complement is sent off to the interface (i.e., PF and LF) before H probes for its goal. Spec,vP and Spec,ApplP are subject to search as they are at the edge of a phase. I adopt Chomsky’s (2000; 2001) probe-goal system based on (un)interpretable features. H has the uninterpretable feature [uh], which is the probe, and the goal DP has the interpretable feature, which can be specified as [+h] or [–h]. It is possible to entertain a different style of agreement mechanism, but the main point of this work remains the same and is not affected by the subtle differences in how agreement is achieved. The technical tools used for agreement are not the primary focus of this paper.

(8) and (9) illustrate H’s failure of agreement with the DO, which is spelled out prior to the introduction of H. v spells out its complement, VP, containing the DO, and Appl spells out its complement, namely the DO. As a result, there is no step in the derivation where H can probe for the DO in syntax.

The same explanation applies to other constructions observed in the language. Pylkkänen (2008) claims that some applied arguments are born higher than others. Beneficiaries and causees are commonly taken to be applied arguments introduced high in the structure. McGinnis (2000) provides evidence that both high and low ApplPs can be phases. According to Kim (2011), Lee & Nie (2022), and Lee (2023; 2024), Korean exhibits high and low applied arguments. Based on this view, coupled with our assumption, all applied arguments are predicted to be eligible for h case assignment. This is because an applied argument, regardless of its syntactic height, sits at the edge of a phase and can participate in H agreement. (10) and (11) show that this prediction is borne out.

Overall, the subject together with the applied arguments are assigned an h case marker in a parallel fashion. Once again, the DO is excluded from this alliance.3

A reviewer points out that Korean exhibits verbs that are sensitive to the honorific status of the DO. In (12) and (13), for example, the predicates poi(p) ‘to meet’ and mosi ‘to take care of’ are only compatible with the honorified objects, halapeci ‘grandfather’ and sensayngnim ‘a teacher’. Replacing these nominal arguments with a non-honorifiable argument such as ai ‘a child’ devastates the derivation, suggesting that the verbs are sensitive to object honorification.

In cases where a non-honorifiable argument such as ai ‘a child’ participates in the derivation, a different form of the predicate has to be used as shown in (14) and (15).

Under the current analysis based on phases, a verb may be sensitive to the honorific status of the DO even though the DO itself cannot bear an honorific case marker. The following PF Vocabulary Insertion (VI) rules for the predicates mentioned in (12)–(15) help us address this puzzle.

The morphologically marked forms poi and mosi in (16a) and (17a) are realized when a syntactic copy of the honorified DO sits in the complement position of V inside VP. VP is basically the local environment in which allomorphy is conditioned. In all other contexts, the verbs take on their default forms, manna and tolpo, as shown in (16b) and (17b). Crucially, the VI rules in PF apply even if agreement between H and the DO is unsuccessful in narrow syntax. VI and agreement are two independent operations taking place in two separate workspaces. Thus, it is possible for a predicate to take on a DO-sensitive honorified form even in the absence of an h case marker on the DO.

The analysis advanced in this work makes predictions about the left periphery. In order to house HP, the structure must contain a phase. CPs are phases (Chomsky 2000; 2001), which suggests that CPs are selected by H. This, in turn, predicts that an argument at the edge of CP can be assigned an h case marker. The addressee is assumed to be represented in the left periphery (Speas & Tenny 2003; Haegeman & Hill 2013, among many others). (18) and (19) host an overt addressee, and (19), in particular, is an addition to the empirical coverage of the current analysis based on phases.

(20) demonstrates how agreement is implemented between H and the addressee.

A central claim of this work is that honorific markers are sensitive to phases. Independent support for a phase-based treatment of honorific morphology comes from Choi & Harley (2019). They argue that honorific markers are realized through a mechanism known as node-sprouting (also known as dissociated morpheme insertion) in Korean. Based on their approach, node-sprouting repairs a syntactically well-formed but morphologically incomplete structure by inserting a morpheme into the configuration. Crucially, Choi & Harley propose that node-sprouting is phase-sensitive, in that it applies once per phase. In the following example, the honorific marker -si is realized twice: one on the inner v associated with ka ‘to go’ and the other on the outer v associated with ha ‘to do’.

The present analysis lends support to this view by showing that phasehood plays a key role in determining the realization of honorific features. The discussion offered in this work thus sets the stage for a broader inquiry into why honorific features should be a part of grammar and how they are handled in syntax.

The empirical landscape of Korean honorific case marking reveals a consistent pattern: while subjects and IOs freely allow h case morphology, DOs do not. The puzzle revolves around the fact that no h case marker is ever associated with the DO. The absence of h.nom on DOs is used as evidence to suggest that morphological syncretism alone cannot handle the full empirical picture. I have argued that the primary phenomenon of interest is closely associated with an independently motivated syntactic concept, namely phases. By adopting a phase-based analysis where vP, ApplP, and CP constitute domains of agreement and by assuming that H probes for a [+human] argument at the phase edge, we correctly account for the absence of an h case marker on DOs. DOs fail to receive an h case marker not because of syncretism but because they are structurally unable to occupy the position necessary for agreement with H. We have also seen that the core insights of this proposal apply elsewhere in the grammar.