Korean allows null arguments in finite clauses. According to pro analyses of null arguments, null arguments are silent forms of overt pronouns. Unlike sloppy interpretations of null arguments, those of overt pronouns are somewhat restricted and show speakers' variation. Accordingly, the presence of sloppy interpretation of null arguments has been considered as weak points of the strict pro analyses. In this paper, we examine two factors that may affect sloppy interpretation of overt pronouns: singular vs. plural on the one hand, and human vs. nonhuman on the other. In addition, we discuss a non-trivial empirical advantage our pro analysis has in comparison with another type of pro analysis such as NP Substitution analysis.

A tree model that is typically used in data mining and machine learning has been recently adopted to analyze multifactorial data in corpus linguistics. However, Gries (2019) claims that tree-based approaches fail to deal with a data set in which not all input variables are necessary for predicting the target variable accurately. He demonstrates that a tree model reports a three-way interaction even when a two-way interaction can predict the target variable with 100% accuracy because there is another third significant input variable. In this article, I will argue that the result of a conditional inference tree applied to such a data set suggests that a researcher should analyze a two-way interaction effect. In other words, the result allows two different interpretations that are related to the structure of a data set and requires to test whether a two-way interaction is sufficient to explain the data. In doing so, I will show that the presence of the third input variable does not hinder the detection of a perfect two-way interaction that is displayed by a symmetric tree figure.

Korean uses grammatical markers to express evidentiality which refers to the source of information of an event indicating whether speakers directly witness the event, indirectly infer the source from general knowledge or perceptual cues, or are told about the event. Many studies have explored linguistic systems for Korean evidential types. However, little has been investigated about how well speakers' intention on expressing the source of information has delivered to listeners. In this study, we attempted to quantify the degree of listeners' perceptual sensitivity on speakers' linguistic cues representing whether speakers witness expressed events. Manipulating eight sentence-ending markers (i.e., -e, -ta, -ci, -canha, -ney, -te-, -kwuna, -tay) of event statements, we asked Seoul and Yanbian dialect users to rate the likelihood of speakers' witnessing expressed events from 1 (least likely) to 5 (most likely). Our results showed that event statements ended with -e marker were scored the highest rates (speakers were highly likely to witness the events), whereas event statements ended with -tay marker were scored the lowest rates (speakers were least likely to witness the events). We discuss that the source of information cued by grammatical markers is not always perceived to listeners in a way that the marker grammar defines.