Using 129 natural conversations and 185 episodes of television drama conversations as well as the theoretical frameworks of usage-based theory and grammaticalization, I investigate two forms of ‘or’ in Korean, tunci and kena. Generally believed to be largely interchangeable, these two forms’ actual usages have never been compared. I demonstrate that the two are selectively used in conversation, and propose that three types of factor influence the selection. The first factors are genre and setting. In formal settings and formal descriptive writing, kena is preferred. The second factor is the presence of negative markers: while 43% of kena tokens were used in negative sentences, only one token of tunci was. The third factor is the distinction between ‘propositive’ tunci and ‘descriptive’ kena. In conversation, tunci occurs mainly in making suggestions and presenting plans, and hence is oriented toward the future tense and the irrealis domain. Kena occurs mainly in describing present and past situations, facts, or premises, and hence is oriented toward the present and past tense and the realis domain. This study confirms the usage-based theory’s argument that speakers possess extensive procedural knowledge of language: Korean speakers may not be conscious of it, but they differentiate tunci and kena with high regularity. Lastly, I explain their differences based on their history. The older form, kena, dates back to the 8th century, whereas tunci became widespread only in the 19th century. Because tunci retains its original semantics of probability, tunci tends to occur in presenting suggestions and plans (irrealis) and not in factual descriptions.

我使用129个自然对话和185个电视剧对话情节，以及基于用法的理论和语法化的理论框架，研究了韩语中的两种“或”形式，即tunci和kena。通常认为这两种形式在很大程度上可以互换，但从未对这两种形式的实际用法进行过比较。我演示了在对话中有选择地使用这两种方法，并提出了三种类型的因素会影响选择。第一个因素是类型和背景。在正式场合和正式的描述性写作中，kena是首选。第二个因素是否定标记的存在：虽然43％的kena标记用于否定句中，但只有一个tunci标记曾是。第三个因素是“命中性” tunci和“描述性” kena之间的区别。在对话中，tunci主要发生在提出建议和提出计划中，因此面向未来时态和irrealis域。Kena主要发生在描述当前和过去的情况，事实或前提，因此面向当前和过去的时态和领域领域。这项研究证实了基于用法的理论的论点，即说话者拥有广泛的语言程序知识：朝鲜语者可能没有意识到这一点，但他们区分了tunci和kena规律性高。最后，我根据他们的历史来解释他们之间的差异。较早的形式kena可以追溯到8世纪，而tunci只是在19世纪才开始流行。由于tunci保留了其原始的概率语义，因此tunci往往会在提出建议和计划时出现（无意识），而不是在事实描述中出现。