Using the data-based methods of language research, this article addresses whether the seltukha-constructions (persuade-constructions) in Korean indeed undergo a control phenomenon as generally assumed and argued in the literature. As is well known, there are two essential properties of controls across languages; (i) the controllee should be co-indexed with another element in the sentence and (ii) the controllee should be silent. Focusing on these two properties, the present study provides multi-pronged data taken from corpus exploration, context-sensitive survey, and language experiment. The empirical investigation casts a doubt on the control analysis of seltukha-constructions; in particular, it is borne out that the embedded subject (controllee) of the seltukha-constructions can appear explicitly, and the embedded subject – whether it be silent or not – is not necessarily required to be co-indexed with another element in the sentence. These lead us to the conclusion that the seltukha-constructions can be accounted for by means of a pro-drop rather than a control. One remaining issue in the analysis is such that the seltukha-constructions may sound awkward if the two NPs (matrix object and embedded subject) refer to the same individual and they appear simultaneously. To account for the awkwardness, we propose the Anti-redundancy Hypothesis; i.e., two NPs referring to the same entity or having the same form tend not to appear right next to each other. The current experiment further demonstrates that the hypothesis works as a general tendency rather than a stipulation about seltukha-constructions. Building upon the empirical data, we propose that the theory of controls in Korean (and hopefully other languages) should be reexamined from the bottom up.
1. Introduction
2. Basic data
3. Literature review
4. Corpus analysis
5. Co-indexation
6. Experimental analysis
7. Discussion
8. Conclusion