Obligatory Split Control into Exhortative Complements in Korean

I argue that in Korean, split control into exhortative complements is in fact obligatory control (OC). This runs contrary to most analyses, as split control in English is usually characterized as a form of non-obligatory control (NOC) (e.g., Williams 1980, Hornstein 1999). The fact that Korean exhibits split OC suggests two possible conclusions: (a) the typology of control is not uniform across languages and what is OC in one language may not be in another, or (b) split OC is a true form of OC and what defines split OC is poorly understood. In this squib, I argue for the second possibility.¹

In section 1, I provide a brief outline of control in Korean and empirically detail split OC in English as well as in Korean. In section 2, I examine the implications of the Korean data for two current theories of control, OC as A-movement, beginning with Hornstein 1999, and OC as Agree, which began with Landau 1999. I show that these approaches cannot account for the data in a straightforward manner. In section 3, I provide the beginnings of a possible alternative analysis.

1 Split Control in Korean Is Obligatory Control

1.1 Control Contexts in Korean

Since at least the appearance of Yang 1984, it has been observed that control in Korean arises from the combination of a matrix predicate capable of control and an embedded mood marker that triggers the control reading of the verb. I refer to this combination as a control context. Consider the examples in (1a–b).

1. 1.

   1. a. Inho₁-ka  Jwuhi₂-eykey [PRO_*1/2 cip-ey

      Inho-NOM Jwuhi-DAT   home-LOC

      ka-la]-ko mal-ha-yess-ta.

      go-IMP-C tell-do-PST-DECL

      'Inho told Jwuhi to (Jwuhi only) go home.'

   2. b. Inho₁-ka  Jwuhi₂-eykey [pro_1/2/3 cip-ey

      Inho-NOM Jwuhi-DAT   home-LOC

      ka-n-ta]-ko  mal-ha-yess-ta.

      go-IND-DECL-C tell-do-PST-DECL

      'Inho told Jwuhi that he/she/someone is going home.'

In (1a), the imperative mood marker la creates a control context in conjunction with the matrix control predicate mal-ha-ta 'to tell'. If the embedded mood marker is changed to the indicative as in (1b), the control reading is lost and a pro reading becomes possible. Other characteristics of control in Korean are that tense markers cannot be used in the complement clause and that long-distance control, arbitrary readings, strict readings, and de re readings of PRO are all impossible (Madigan 2006a).

1.2 Split Control in English

It has long been assumed that in English, split control is a form of NOC and is a property that negatively defines OC; in other words, if a control verb does not allow split antecedence, then it is an OC verb. For instance, one supposedly cannot say (2).

1. 2. *Jim₁ promised Mary₂ PRO₁₊₂ to go home together/with each other.²

While most authors take this as an inescapable truism, some do recognize it as problematic (Landau 2001). The truth is that split OC does exist in English, however marginal or variable among speakers it may be. For example, many speakers, including myself, find (3a–c) perfectly acceptable.

1. 3.

   1. a. Jim proposed to Mary to go to the movies by themselves/ with each other.

   2. b. Tim suggested to Masahiro to introduce themselves (to each other).

   3. c. John proposed to Mary to help each other.

      (Koster and May 1982)

Given that there are at least some cases of split control in English, the question is whether this control is obligatory or nonobligatory. In what follows, I show that examples like (3a–c) are indeed cases of OC. In doing so, I use the familiar diagnostics in (4), which Landau (2001) argues to be central in defining OC.³

1. 4. Basic properties of obligatory control of PRO

   1. a. Arbitrary control is impossible in OC.

   2. b. Long-distance control is impossible in OC.

   3. c. A strict reading of PRO is impossible in OC.

   4. d. A de re reading of...