A CORPUS BASED EXPLORATION OF THE PROGRESSIVE -KO ISS CONSTRUCTION 
IN L1, L2, AND TEXTBOOK KOREAN 

By 

Steven G. Gagnon 

A DISSERTATION 

Submitted to 
Michigan State University 
in partial fulfillment of the requirements  
for the degree of  

Second Language Studies – Doctor of Philosophy 

2024  

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ABSTRACT 

Due to the typological differences between Korean’s aspect system and English’s aspect system 

in terms of progressive construction -ko iss, learners can no doubt have difficulty acquiring and 

using the -ko iss construction in learner Korean. This dissertation investigates two main points: 

(i) how is the -ko iss construction used in real-world Korean, including L1 Korean and L1 

English and L1 Japanese learner Korean, and (ii) the way -ko iss is taught and used in textbooks 

as a main source of input for learners of Korean. To answer these questions, I use 

collostructional analysis to assess association strengths between verbs and the -ko iss 

(progressive) and simple (non-progressive) constructions to identify verbs that are well-attested 

with L1 and L2 Korean. Finally, I take an exploratory approach to using logistic regression to 

model L1 and L2 Korean data, the results of which can provide some insights into L1-L2 -ko iss 

usage, and insights from this initial regression analysis provide meaningful information to 

improve modeling of L1 and L2 Korean in future studies. The main takeaways from this study 

are:  

(a)  Verbs co-occurring with -ko iss in the written Sejong corpus included a wide variety of 

usage cases, including many instances of stative or mental-type verbs, including al 

(know), mid (believe), among others. 

(b) Verbs co-occurring with -ko iss in the learner data showed a positive sign in that learners 

use and acquire the -ko iss construction’s various semantic meanings, including its use 

with stative verbs. However, semantic domains used with -ko iss are limited when 

compared with the L1 data.   

(c)  In textbooks, a limited number of verbs is introduced with -ko iss at the beginner levels. -

ko iss is also taught in textbooks as a prototypical action in progress progressive 

 
 
 
construction, without clear direction instruction on other senses of -ko iss. Further, across 

both textbook series, the frequency -ko iss is used at is low (maximum around 300 

occurrences in a textbook series). Textbooks incidentally use -ko iss outside of the 

prototypical action in progress usage at later levels, however, frequencies are quite low. 

Findings from this dissertation can be used to inform language pedagogy. The list of verbs co-

occurring with the -ko iss construction from the collostructional analysis provides teachers and 

textbook developers with a list of attested to function with -ko iss across a variety of usages 

beyond action in progress. Plenty of examples are also pulled from the corpus for materials 

developers to reference when designing textbook materials. As the aim of language teachers and 

materials developers is to use data-driven insights to improve teaching materials, exposing 

learners to a variety of verbs within contexts or lexical chunks they appear in via textbooks can 

aid in learning complex constructions in Korean.  

 
 
 
 
 
 
 
 
 
 
 
 
 
 
TABLE OF CONTENTS 

I. INTRODUCTION ....................................................................................................................... 1	

II. LITERATURE REVIEW ........................................................................................................... 4	

III. METHOD ............................................................................................................................... 26	

IV. RESULTS ............................................................................................................................... 40	

V. DISCUSSION ........................................................................................................................ 101	

REFERENCES ........................................................................................................................... 114	

APPENDIX A: DISTINCTIVE COLLEXEME ANALYSIS I .................................................. 120	

APPENDIX B: DISTINCTIVE COLLEXEME ANALYSIS II ................................................ 136	

APPENDIX C: DISTINCTIVE COLLEXEME ANALYSIS III ............................................... 137	

iv 

 
 
 
 
I. INTRODUCTION 

Corpus-driven explorations into real-world language use allow language teachers and researchers 

to uncover language patterns that occur frequently, which allows for development of teaching 

materials that mirror real-life usage rather than a speaker’s language intuition. Led by John 

Sinclair, corpus linguistics and its potential as a proper field in applied linguistics was bolstered 

forward in 1987 with the publication of the COBUILD English Language Dictionary, a project 

which stemmed from the goal of building corpus-driven materials for second language learners 

of English. Such work helped push corpus-based and usage-based approaches forward towards 

recognizing that co-occurrence patterns of lexical items and syntactic structures are inseparably 

intertwined, thus revealing the link between form and meaning, particularly for commonly 

occurring collocation patterns (Sinclair, 2004).  

The move towards corpus-driven, evidence-based approaches with large corpora as the 

backbone have influenced the development of learning materials such as textbooks, dictionaries 

(e.g., COBUILD), and so forth. As Sinclair puts it as the title of his 2004 book, we must “trust 

the text.” In other words, we must rely on real-world language examples in corpora to guide our 

understanding of language, which is a notable deviation from generative approaches to 

linguistics which rely heavily on linguistic intuition (Abbot & Tomasello, 2006). This is of 

particular importance as the use of corpora has allowed researchers to discover patterns which 

have otherwise gone unnoticed (e.g., Sinclair, 1997) when relying solely on a researcher’s own 

intuition.   

Since the seminal work on the COBUILD project, a multitude of corpus-based and 

corpus-driven learner materials such as dictionaries have been developed, particularly in the case 

of English as a second language teaching. However, despite the continued rise in corpus-based 

1 

 
studies and publications, there is much work to be done by corpus linguists to draw meaningful 

connections between research, teaching practices, and educational materials. In fact, even in the 

case of English language teachers, Römer (2011) argues that “the practice of English Language 

Teaching (ELT)… seems to be only marginally affected by the advances of corpus research” (p. 

206). The sentiment was borne out in results in an earlier study by Römer (2005) which revealed 

that real-world English indeed differs from the English found in language teaching materials. 

Similarly, corpus-based explorations of Korean language learning materials (e.g., Jung, 2022) 

have shown deviations between language use in real-life and the language that appears in 

textbooks.  

I agree with Römer that in the field of corpus linguistics “much work still remains to be 

done in bridging the gap between research and practice” (Römer, 2011, p. 206). 

To that end, this dissertation outlines a corpus-driven approach to identify language patterns in 

L1, L2, and Textbook Korean with the goal of illuminating key differences that arise in learner 

language and offering suggestions for the creation of teaching materials such as language 

textbooks. Following in the footsteps of previous work on corpus data and textbook analysis 

(e.g., Jung, 2022), and Römer, 2005)), and using robust statistical methods to tease apart the 

differences between L1, L2, and language materials, the proposed dissertation project will be 

conducted with the goal of identifying verbs with co-occur with the progressive in L1 and L2 

Korean, what factors predict the choice of a progressive form, and how these usage cases 

compare with Korean language textbook materials. This project will help move the field of 

Korean second language acquisition forward and help close the gap between teachers and 

researchers. Ultimately, this project addresses the needs of stakeholders in language education, 

2 

 
namely students, teachers, and language education material developers to provide learners with 

robust materials for language acquisition. 

3 

 
 
 
II. LITERATURE REVIEW 

This study has two main aims which both work towards (i) understanding and modeling usage of 

the progressive and simple form (i.e., the “(non-)progressive”) both within and across language 

varieties of L1 and L2 Korean, and (ii) comparing L1 use of the progressive with language 

appearing in textbooks. Association strengths between verbs and the progressive and non-

progressive are assessed using collostructional analysis (Gries & Stefanoswitsch, 2004) a 

statistical method to assess the attraction of a verb to a particular construction. Following that, I 

dig deeper to identify what predictors may impact the choice to use to use a progressive versus a 

non-progressive (e.g., aktionsart category, semantic domain, and L1) using logistic regression. 

To set the scene, studies on corpora and textbooks, L1 and L2 Korean and the progressive 

construction in Korean (and other languages), and descriptions of the progressive in Korean are 

provided.  

2.1 Corpora and textbooks 

In nearly all language education contexts, textbooks provide an important source of input for 

language learners. Textbooks provide many benefits to learners: they allow for concentrated 

practice on vocabulary, grammatical forms and functions, and practice activities for learners on 

the one hand, while simultaneously easing the burden on the language teacher by providing some 

materials to use in the classroom or for homework. When designed well, textbooks can provide 

learners with the chance to practice their target language (Lam, 2009) even if they do not have 

ample opportunities to communicate with L1 and expert interlocutors. As such, stakeholders in 

language education (including not only language teachers, but also textbook developers and 

publishers, and even students) should be invested in the research and development of language 

textbooks. 

4 

 
While the value of textbooks as learning resources is undeniable, there is a need for more 

robust explorations and comparisons of features in corpora and comparing them with how they 

are presented in textbooks and materials. However, insights from robust corpus studies seem to 

not be implemented often, and this may mean that certain criticisms of textbooks, namely that 

they often do not reflect the way language is used in the real-world due to contrived examples for 

the sake of grammar instruction but at the expense of authentic and meaningful input (e.g., 

Timmis, 2014). It may be for these very reasons that the usefulness of learner corpus research 

and its application to teaching and materials development in general has been called into 

question. For example, Flowerdew (1998) stated that “…the implications for pedagogy are not 

developed in any great detail with the consequence that the findings have had little influence 

on… syllabus and materials design” (p. 550). Römer (2006) also commented on this issue saying 

that while the value of using corpora to inform materials development has “obvious and 

recognized strengths… it seems that there is still a strong resistance towards corpora from the 

side of students, teachers, and materials writers” (p. 124). Further, when insights from corpora 

are used in textbook development, it seems that there are still weaknesses in their 

implementation. Timmis (2013) highlights this in a chapter on developing materials citing 

Koprowski (2005) who noted that while textbook and materials designers are open to 

incorporating language chunks and multi-word units in their materials, their selection is often not 

informed by corpora and still largely relies on the developers’ own sense and intuition. 

Koprowski also notes that the diversity of multi-word units selected for materials (in the case of 

English language materials) often place excessive emphasis on simple collocations rather than, 

say, phrasal verbs or longer multi-word expressions which could be identified in L1 corpora. 

Römer (2006x) also comments on mismatches between language found in corpora and materials, 

5 

 
 
stating that: “For all items investigated, researchers found considerable mismatches between 

naturally-occurring English and the English that is put forward as a model in pedagogical 

descriptions” (p. 126).  

Research and development of language textbooks using corpus-based methodologies can 

provide stakeholders with insights to assess the language that appears in textbooks and how well 

that language reflects language use in real life. In an aptly titled paper, Corpora and Language 

Teaching: Just a Fling or Wedding Bells? on the state of corpus-based research and the field of 

language teaching where she equates corpus-based research and language teaching as either in a 

fling or on the verge of great collaboration (wedding bells), Gabrielatos (2005) highlights how 

corpora can be leveraged to create meaningful outcomes, such as developing learning and 

teaching materials, and examining textbooks to identify (i) what language forms learners are 

exposed to and (ii) facilitate the research and development of textbooks and other materials or 

assessments. This is illustrated in the figure borrowed from Gabrielatos (2005), which shows the 

potential to leverage corpora in textbook development. To summarize the main points, L1 

corpora can be used to identify real-world usage of language, and this can be compared with both 

L2 corpora and textbook corpora. This is especially critical as textbooks have been found to not 

be an accurate portrayal of L1 language use (e.g., Römer 2004, 2005), sometimes due in part to 

language as it dynamically shifts and changes over time. 

6 

 
 
 
 
 
 
Figure 1.1. 

Corpora and ELT (copied from Gabrielatos’ 2005 article Corpora and Language Teaching: Just 

a Fling or Wedding Bells?). 

Such analyses can also uncover trends in changes in language and how they are (or are 

not) represented in learning materials. A recent study by Belli (2018) investigated stative verbs 

in the progressive aspect in English language textbooks. This is key for textbook developers as 

the progressive in English has been undergoing a shift in meaning to include stative readings. 

Traditionally, stative verbs in English such as love, want, feel, etc., “have been known [as] the 

verbs which cannot or rarely occur in the progressive form as evidenced in a number of 

previously written English textbooks” (Belli, 2008, p. 2018). In fact, some previous textbooks 

have even claimed that stative verbs are incompatible with the progressive (e.g., Anderwald, 

2012 as cited in Belli, 2008). However, such extreme statements are challenged by the fact that 

stative verbs do appear in the progressive form (e.g., see Granath & Wherrity, 2014). In the case 

of Belli’s (2008) study, the textbooks under investigation were “corpus-informed,” meaning that 

they “were designed by authors who made use of various native English corpuses [corpora], 

7 

 
 
  
which reflected the target language as it is currently written and spoken” (p. 126). Corpora used 

in the design of the textbooks included the Cambridge International Corpus and the Cambridge 

Learner Corpus, The Corpus of Contemporary American English (COCA), among others. The 

aim of the study was to identify how stative verbs in the progressive aspect were incorporated 

into corpus-informed textbooks. An interesting finding in this study was that stative verbs which 

can be associated with the progressive in English, particularly verbs expression emotion (e.g., 

want, love, feel) were in fact included in the corpus-informed textbooks with the progressive 

form, in contrast to previous textbooks that include descriptions of stative verbs being 

ungrammatical with progressive forms. This is also in-line with studies such as Freund (2016) 

which show that “…certain statives attract progressive aspect in particular contexts, while others 

remain resistant to it…” (p. 59), and that certain verbs may have increased in their usage with a 

progressive (in colloquial British English), a nuanced insight which can inform textbook 

development. In the case of the corpus-informed textbooks, Belli notes that the usage cases for 

stative progressives in English, such as referring to a situation as dynamic, were in fact explained 

in the corpus-informed textbooks. 

Another textbook analysis by Jung (2022) investigated Korean language textbooks and 

how postpositions (specifically, particles such as -ey or -eyseo) were incorporated into the 

textbooks. Traditionally, certain Korean postpositions are introduced in textbooks with functions 

that express either static place or dynamic location (e.g., Jeong, 2011; Kim, 2011). As Jung 

(2022) notes, “this dichotomy of postpositions that share similar functions in textbooks may 

confuse language learners when they are exposed to the natural language use environment, which 

is not consistent with what is presented in the textbooks.” (p. 202). In Jung’s own analysis, she 

investigated postpositions by addressing their frequency of occurrence, checking for commonly 

8 

 
 
co-occurring verbs, and keyness analysis. The corpora employed in that study were the Sejong 

corpus (written and spoken) as an L1 reference, and a corpus of two series of Korean textbooks. 

For the textbook corpus, data were compiled from 16 volumes in total and covered four 

proficiency levels (typical of what might be used in a four-year Korean program at a university). 

The analysis revealed that as the textbooks’ level (beginner through advanced) increased so too 

did the number and variety of verbs co-occurring with postpositions. Jung was also able to 

uncover that certain verbs which commonly occurred with post-positions in L1 Korean were 

lacking representation in the textbooks. Also of note is that the “location, position, and 

existence” function when a postposition is used with the predicate -iss, to exist, was largely 

lacking in the textbook corpus. Jung notes that this is in line with previous work where learners 

exhibited lower accuracy with location functions and higher accuracy with direction functions of 

constructions with the postposition -ey (e.g., Kim & Guo, 2016). For the purposes of this 

dissertation, this recent example of a study of L1 and textbook Korean shows how a comparison 

of (a) L1 corpora and (b) language learning materials can provide (i) insights into how real-world 

and textbook language differ and (ii) highlight areas where language learning materials can be 

modified and improved. 

2.2. Textbooks and corpora – the way forward and towards incorporating robust analyses 

As discussed above, corpora can be a powerful tool when creating and designing textbooks and 

materials for language learning. However, that is not to say that teacher-researchers and 

materials developers need to put intuition to the wayside when designing textbooks or preparing 

lesson plans. I believe Timmis (2014) aptly put it when suggesting the term corpus-referred 

materials (as opposed to corpus-based or corpus-informed) may be the way forward. Timmis 

states that:  

9 

 
“A corpus-referred approach, I would argue, explicitly allows an honorable place for intuition, 

experience, local need, cultural appropriacy and pedagogic convenience in determining syllabus 

content and the order in which items are taught.” (Timmis, 2014, p. 470). 

Taking a corpus-referred approach allows stakeholders to consider the data that can be 

gleaned from corpora while also taking note of what intuition tells us or what has been shown to 

work in the classroom. For example, in this view, there is not necessarily a need to present 

grammatical structures based on their frequency in a corpus, especially if such features could be 

difficult for learners or require some scaffolding of perceivably simple features in advance (e.g., 

Biber & Conrad, 2010).  

Additionally, there is mounting evidence that input from textbooks can have a positive 

impact on language acquisition. A study by Northbrook and Conklin (2019) investigated whether 

students were able to lexical bundles appearing in their textbooks faster than others. They found 

that, in fact, the input learners received from learning materials including textbooks led to 

students being able to process the lexical bundles they encountered in their textbooks faster. And, 

as the students who participated were lower in proficiency, their study also provided evidence for 

the effectives of input from textbooks even for learners at lower levels. Relating to the present 

study, analyzing potential disparities between L1 and textbook representation can open the door 

to suggest revamping the representation of certain linguistic features in textbooks.  

I also point out that the field of corpus linguistics has been rapidly evolving, and since the 

publication of some aforementioned studies, robust statistical techniques have started to take 

center stage in research conducted in the corpus domain. For example, while traditional corpus 

enquiries into language focused on evaluating frequency counts running keyness analyses, it is 

becoming more common to involve advanced statistical techniques to corpus data. I specifically 

10 

 
 
refer to the use of regression analysis in the field of corpus linguistics as put forward by Gries 

and Deshors (2014). While a complete review of the paper is beyond the scope of this 

dissertation, key points will be highlighted here. In their paper, Gries and Deshors outline how 

historically corpus studies on interlanguage between L1 and L2s have relied on raw frequency 

counts from often comparable corpora. However, there are clear weaknesses when only 

considering frequency counts void of context to account for over- and underuse of linguistic 

features by learners when compared with L1 speakers. Thus, rather than relying on frequency 

counts of certain features when L1 and L2 speakers, for example, write essays about the morality 

of smoking, we can consider the context as linguistic/contextual features (p. 114, emphasis 

added). To quote Gries and Deshors: 

“…we should look at NSs’ choices of can versus may when the subject is animate, singular, 

when the clause is interrogative,… and then compare this to NNSs’ choices of can versus may 

when the subject is animate, singular, when the clause is interrogative. In this view, ‘comparable 

situation’ is now defined much more comprehensively in terms of linguistic/contextual 

features… give way to what we think should be one of the fundamental questions of SLA/FLA 

research: ‘in a situation, S, characterized by features F1-n that the learner is now in, what would a 

native speaker do (and is that what the learner did do)?’” (pp. 113-141).  

Using robust statistical methods (e.g., regression modelling, generalized linear mixed 

effects modeling, or collostructional analysis) can help teachers and researchers reveal what 

makes an L2 speaker’s speech sound markedly different from an L1 speaker despite target 

grammar and vocabulary usage being largely correct. A nuanced approach is that over/underuse 

of features contribute to “foreign-soundingness even in the absence of downright errors” 

(Granger, 2004, p. 132). 

11 

 
As a construction which is known to be difficult for students to acquire, the progressive 

has been the focus of many studies. In-line with the robust statistical methods mentioned above, 

recently, several scholars have begun using robust corpus-based and corpus-driven 

methodologies and advanced statistical techniques, such as regression modeling, generalized 

linear mixed effects modeling, and collostructional analysis (e.g., Römer, 2005; Kranich, 2010; 

Hundt & Vogel, 2011; Rautionaho, 2014; Deshors & Rautionaho, 2018; Fuchs & Werner, 2018; 

Rautionaho, 2020) to measure the attraction of words to certain syntactic constructions or to each 

other within a construction (e.g., Stefanowitsch & Gries, 2003). However, to date, much of the 

work done on the progressive (and indeed, in much of the field of corpus linguistics) has focused 

heavily on learner English and world Englishes (e.g., Römer, 2005; Rautionaho, 2014, 2020, 

among others). While this is no doubt due to the widespread usage of English as a world 

language and a lingua franca which lead to a natural need for corpus based educational materials 

and dictionaries (e.g., COBUILD), a gap in the literature exists when it comes to other 

languages.  

To summarize, corpus linguists have a great opportunity to serve in the role of textbook 

and learning materials development, particularly when evaluating what types of forms and 

functions should be included in materials. To that end, robust corpus-based analyses can arm us 

with data on constructions, their semantic meanings, and usage cases (e.g., Gries et al. 2005) as 

usage data pulled from corpora offer clear insights into a construction (and lexical items that 

appear in it as well as semantic descriptions). As has been shown in the aforementioned studies, 

corpus-based investigations can (i) reveal gaps between real-world and textbook language, (ii) 

identify changes in language over time, and (iii) aid in the improvement of learning materials 

that better match natural language. In the next section, I will discuss corpus-based work on the 

12 

 
 
progressive. I will discuss research on the acquisition and usage of the progressive in learner 

language and highlight key aspects of the progressive such as its use with stative verbs in both 

English and Korean. I will touch on statistical methods used in corpus linguistics to address the 

progressive, which will lead into the present study’s methodological considerations.  

2.3. Theoretical underpinnings: usage-based approaches to second language acquisition 

In usage-based viewpoints of language acquisition, language learning is driven by previous 

experiences with language, and these repeated exposures and experiences over time result in the 

cumulative frequency effects necessary for uptake of linguistic constructions. Put simply, 

language acquisition happens after repeated exposure as language learners subconsciously tally 

up co-occurrence rates of forms with functions, which over time become entrenched and 

automatized (e.g., Bybee, 2013; Tomasello, 2003). As frequency effects and exposure to 

linguistic features are key to the automatization of constructions in usage-based viewpoints, 

studying both how learners use linguistic constructions and the input they are exposed to can 

provide a workable framework for teachers and materials developers in their pedagogical 

practices. 

Thus, to analyze the choice of the progressive construction or the non-progressive 

construction in both L1, L2, and textbook Korean, I take a usage-based approach and consider 

that frequency effects drive language acquisition for both L1 and L2 speakers. In other words, 

the more often a speaker encounters a certain word, construction, or collocation, the more 

entrenched that piece of language becomes as learners are sensitive to the exposure patterns 

developing probabilistic knowledge (e.g., Ellis, 2002; Ellis 2008), and these repeated exposures 

which lead to entrenchment of language form-function mappings are as important, if not more 

so, than conscious noticing and becoming aware of form-function mappings (Schmidt, 1990). 

13 

 
Thus, it is important to identify which verbs co-occur with the progressive in textbooks and how 

these co-occurrence patterns differ from real-world usage, which the aim of identifying ways to 

improve the textbooks, which serve as a main source of input, for learners. That is to say, while 

L1 influence can certainly play a role in a learner’s acquisition of a linguistic form, ensuring that 

their input (such as textbooks) matches real-world language as much as is realistically possible 

can propel learners to notice and acquire constructions in their L2.  

In addition to frequency effects, a learner’s L1 can also influence the uptake and 

acquisition of a linguistic feature. For example, existing literature on progressive and continuous 

aspect constructions shows evidence for variation between L1 and L2 usage, and that L2 usage 

may be influenced by interlanguage effects from the L1. In a corpus-based study of 

argumentative essays from the ICLE (International Corpus of Learner English), Virtanen (1996) 

found that differences in a learner’s L1 lead learners to use the progressive construction in 

different amounts. A comparison of essay data from the L1 Finnish, L1 Finland-Swedish (a 

dialect of Swedish spoken in Finland), and L1 Swedish revealed statistically significant 

differences in the usage rates of the progressive. Virtanen noted that the rate at which a learner 

used a progressive in their writing differed depending on their L1. Notably, L1 Finnish learners 

used the progressive significantly less in their writing than the other two learner groups (L1 

Finnish-Swedish and L1 Swedish). Virtanen attributed this difference in usage to L1 influence, 

stating that students’ usage of the progressive “seems to vary according to their mother tongue 

background” (p. 301). 

When considering frequency effects on the uptake of form-function mappings, it is 

possible to tease apart more detail than simply a verb’s association with a particular construction. 

Within the domain of usage-based explorations on interlanguage, variationist approaches are 

14 

 
useful when analyzing what predictors, such as the lexical aspect or semantic domain, of the verb 

in question will gear a speaker towards the choice of one variant over another variant. For 

example, Deshors (2011) and later Gries and Deshors (2014) exhibited how the variation 

between the choice to use may or can in English interlanguage can be explained with several 

predictors such as speaker (e.g., L1/L2), form (may/can), and subject animacy, among others. 

The analysis showed that certain grammatical features (e.g., aspect and negation) can lead L2 

speakers of English to use may and can in different ways.  

For the progressive specifically, recent studies have taken usage-based and variationist 

approaches to explore patterns in a speaker’s choice of the progressive and non-progressive (e.g., 

Deshors & Rautionaho, 2018; Fuchs & Werner, 2018; Hundt & Vogel, 2011; Kranich, 2010; 

Rautionaho, 2014; Rautionaho, 2020; Römer, 2005). In the case of English, corpus-based 

variationist explorations have shown that the progressive construction may most often be chosen 

with verbs in the present tense, verbs which are dynamic, and when the subject is animate, as 

revealed by multifactorial analyses by Hundt, Rautionaho, and Strobl, 2020, and Rautionaho et 

al. 2018. Specifically, Hundt and colleagues identified that tense, modality, verb type, and 

animacy of the subject were all important predictors in the choice of a progressive or non-

progressive. More specifically, they revealed that dynamic verbs and the present tense were 

significant predictors of the choice to use the progressive aspect in the corpus data. Rautionaho 

and Hundt (2021) considered the context in which the progressive occurred in their data (the 

International Corpus of English) and found that in addition to durative situations calling for the 

progressive (e.g.: consider, dance, and stay, p. 616), having a progressive appear in the 

preceding context also lead to an increased usage of the progressive through syntactic priming. A 

preceding study conducted by Deshors and Rautionaho (2018) explored the variation between 

15 

 
 
the choice of the progressive and non-progressive construction based on semantic domain and 

lexical aspect category (aktionsart) of the verb (among other categories) and found that “more 

often than not, writers’ constructional choices are not influenced by a single linguistic factor… 

but rather by the combined influence of (or the interaction between) two factors…” (p. 238). 

Their multivariate analysis showed how semantic domain plays a role in the choice to use, or not 

use, a progressive construction, as it was the only annotated feature which did not have an 

interaction effect with other features. Thus, we can conclude that multivariate usage-based and 

variationist explorations of patterns in the progressive and non-progressive should include the 

semantic domain of the verb as it appears to be a significant factor regardless of variety of the 

speaker (e.g., L1/L2), lexical category of the verb, or genre. 

In particular, when considering data from a usage-based perspective with a focus on 

variation between a choice of construction a or construction b, it is important to consider 

constructions which are as functionally and semantically similar as possible. That is to say, we 

can investigate alternation when the choice of either construction is possible. Rautionaho and 

Hudnt (2021) also point this out and note that “what allows us to treat progressives as part of an 

alternation is that we carefully limit our dataset to instances where both variants are a potential 

choice” (p. 602). In short, to ensure that variation between the choice of progressive and non-

progressive is represented, studies such as Rautionaho and Deshors (2018), Hundt et al. (2020), 

Rautionaho (2020), Rautionaho and Hundt (2021), extract set amounts of exemplars appearing in 

the progressive and the non-progressive. Post-extraction, exemplars are randomized and 

manually checked to be included or excluded based on certain criteria. By following strict 

criteria, this allows researchers to identify what factors can impact the choice of a progressive or 

non-progressive in a variationist approach. Namely, all verbs in the aforementioned studies only 

16 

 
 
included verbs which could appear in both the progressive and the non-progressive. To borrow 

Hundt et al.’s (2020) example, it is a difference that can be seen in sentence variations like he 

was driving along the road and he drove along the road (p. 82). As Rautionaho & Deshors 

(2018) put it: “To strengthen our analysis, we further limited the data to only include such lexical 

verbs that both occur with progressive and non-progressive constructions in our data set” (p. 

232). Thus, through careful extraction and selective data cleaning, a variationist approach can be 

used to assess what linguistic and contextual factors can influence the choice of a progressive or 

non-progressive construction in different speaker varieties.  

Bringing this discussion back to influence of frequency effects from the input from a 

usage-based perspective, this dissertation also incorporates textbook data to determine potential 

effects of input (with a focus on the verbs appearing with -ko iss) on the usage of the Korean 

progressive in learner language, and, to identify potential gaps in the textbook language which 

may need to be addressed considering the verbs appearing in the progressive in L1 data. This is 

of particular note as input and the frequency at which input occurs plays a major role in the 

uptake of form-function associations. For learners, textbooks are one of the major sources of 

input (e.g.: Römer, 2004) and so textbook language merits investigation as it is one source of 

input that can be controlled, to some extent, to provide learners with useful input for language 

learning. Of course, textbooks can never be considered as holistic or exhaustive representations 

of language for learners, but working towards a more comprehensive representation of language 

in textbooks as it is used in modern Korean is one goal of this study.  

2.4. The -ko iss construction 

In this section, I outline the progressive -ko iss construction in Korean which can express a 

continuous or progressive meaning. Functionally, -Ko iss is comparable to the be… ing 

17 

 
 
progressive construction in English. Key typological differences between Korean and the target 

L2s, English and Japanese, are discussed in terms of potential interlanguage transfer effects as 

well. This will set the stage for discussing the selection of corpora and the statistical methods for 

exploring the use of the continuous and progressive constructions in L1 and L2 varieties of 

Korean. As the main focus of this dissertation is -ko iss, that construction is discussed most in-

depth.  

An in-depth account of Korean grammar is Yeon and Brown’s (2011) book Korean: A 

Comprehensive Grammar. Thus, I use descriptions provided by Yeon and Brown for clarity and 

consistency. Furthermore, their descriptions of Korean are of “the standard Seoul speech in the 

Central dialectal zone” (p. 1) which is most often the target for second language learners of 

Korean.  

To form the progressive, -ko, a suffix, is attached to the base form of a verb, and then iss 

is added after -ko. In writing, there is a space between -ko and iss. While -ko does not change 

form, verb endings (to denote past/present/future tense) honorifics, or conjunctions may be added 

to -iss. This construction is most similar to the English be… ing or the Japanese -te iru for 

denoting an action in progress as can be seen in example (1): 

(1) 미나가지금저녁식사를준비하고있다.

Mina-ka  

jigeum  

jeonyeog sigsa-reul  

junbiha-ko iss-ta. 

Mina-NOM 

now 

dinner-ACC 

prepare-PROG-DECL  

Mina is preparing dinner now. 

However, while the -ko iss form seems similar to the English progressive in example (1), there 

are a few key factors which set it apart from the English progressive. Perhaps the most surprising 

difference between the English and Korean progressive forms is that, unlike in English, the 

18 

 
 
 
Korean progressive is “usually optional and used only for emphasis” (Yeon & Brown, 2011, p. 

214) and “unlike the English progressive or the Japanese -te i… the Korean -ko iss- is not 

obligatory for an ongoing event interpretation (Lee & Kim, 2007, p. 656). A common example of 

this is when someone asks what you are doing, and in Korean, a pragmatically appropriate 

response could take the simple present tense, whereas in English, the progressive is preferred, 

such as in the conversational example (2): 

(2) A: 지금



뭐



해?

Jigeum mwo hae? 

Now  what  do-PRES. 

What are you doing now?  

B: 지금 공부해.

Now 

study-PRES. 

Now, I am studying. 

Another key difference is that the progressive in Korean cannot usually hold a futurate 

meaning. For example, while in English the progressive can be used in the futurate to denote an 

action you are about to do. For example, an English speaker can say I am going now right before 

they depart their location. The equivalent in Korean, in example (3), can only be used to describe 

the action in progress. In other words, one must have already departed their location in order to 

use the progressive -ko iss alongside the verb go: 

(3) 지금 가고 있어요. 

Jigeum ga-ko iss-eo-yo. 

Now  go-PROG-PRES. 

I am going now. 

19 

 
 
 
 
 
 
 
 
 
The Korean progressive construction in question for this study, -ko iss, is interesting 

because it differs in its usage from other languages like English, and even more typologically 

similar languages such as Japanese. For example, the Korean progressive can often be used with 

stative and mental verbs which do not often take the progressive in other languages such as 

English (some verbs that fall into this category include believe, desire, feel, have, know, realize, 

and remember). In particular, a common verb co-occurring with -ko iss is al (know), as can be 

seen in example (4): 

(4) 원경 누나가 이제 그 사실을 알고 있다. 

Wonkyung nuna-NOM ijae geu sasil-eul al-ko iss-ta. 

Older sister Wonkyung-NOM now that fact-ACC know-PROG-PRES-DECL. 

Older sister Wonkyung knows that fact now. 

In terms of what may influence the choice of a progressive or non-progressive with such 

verbs, Lee (2006) says that such verbs “belong to a class of inchoative eventualities which 

describes an instantaneous inception event that starts a continuous state” (p. 697). Thus, a choice 

to use the Korean progressive -ko iss construction with a verb such as know can be because 

progressive aspect in Korean is used with not only actions in progress, but also to states that 

come about due to some event, such as coming to know new information, which would then lead 

a Korean speaker to choose to use the progressive -ko iss construction with the verb know. So, 

another possible translation of (4) above, depending on the context, can include older sister 

Wonkyung is now aware of that fact. 

In the case of psychological or cognitive verbs, such as believe or know, one reason Lee 

points out that may allow the verbs to take a progressive marking is that psychological verbs in 

Korean “…do not have to occur every moment afterwards to maintain their effect” (p. 715). In 

20 

 
other words, once one becomes aware of some fact or situation, for example, the choice to use -

ko iss with know is appropriate as they have entered a continuous state being aware of the fact 

from that point onwards. Given this, it is noted by not only Lee (2006) but other scholars as well 

that the categorization of verbs co-occurring with -ko iss is still up for debate. While in this 

dissertation I explore verbs, such as know, as stative and mental verbs, some scholars categorize 

verbs such as know as accomplishments (e.g., Hong, 1991) or resultative achievements (Ahn, 

1995) since they come about as inchoative events. This issue, to my knowledge, is yet to be 

settled as how best to classify such verbs in Korean. So, I do my best to account for such verbs 

by considering them as stative verbs and mental verbs (as appropriate) but acknowledge a future 

study could account for other categorizations of stative or mental verbs in Korean.   

Continuing with the discussion of -ko iss and typological differences between English 

and Japanese, the Korean progressive construction can be used with imperatives commonly, 

which is a unique feature of the progressive in Korean. While the progressive construction in 

English can be used to give a command or instruction in some situations, this usage may not be 

as common as it is in Korean. For example, as an example of a progressive used with an 

imperative in English, one might say: you are not to be driving late at night. However, the 

semantic difference is that English example indicates an action the speaker intends for the 

listener to do in the future (such as instructing someone to avoid driving late at night going 

forward). In Korean, however, you may use the progressive alongside the verb wait to express to 

your listener you want them to stay at the place they are currently at and to instruct they wait for 

you at that location, and this statement has the same nuance as wait here, directing the listener to 

do the action in the present moment, not in the future. A similar meaning in English could be 

expressed using keep or stay. In real-world usage, imperatives or commands with -ko iss in 

21 

 
Korean are generally used in plain or intimate speech styles rather than formal or honorific 

speech styles. Take note of example (4): 

(5) 여기서 기다리고 있어. 

Yeogiseo gidari-ko iss-eo. 

Here  wait-PROG-PRES. 

Wait here (stay waiting here). 

Finally, and most importantly for the analysis to follow, the Korean simple present and 

the present progressive -ko iss can often be used interchangeably (e.g., I eat my lunch and I am 

eating my lunch are both possible to express an action in progress in Korean), and thus there is a 

need to understand how the Korean progressive is used across L1, L2, and textbook Korean. 

I have provided an overview of the -ko iss construction in Korean. In this study, my 

analysis will focus on -ko iss’s prototypical usage, that is, when it appears sentence final without 

any other connectors or tenses attached. Future studies can explore -ko iss in future/past tenses, 

as well as in conjunctions or negations.  

2.5. Research on continuous aspect constructions in Korean 

There have been several notable studies on continuous aspect constructions in Korean. In this 

section, I highlight some studies on L1 and L2 Korean. As is the case with many learners, 

regardless of the L2, acquiring the progressive construction can be tricky due to differences in 

semantic usage as well as form and function mappings; Korean is no exception.  

Crosslinguistic variation in the usage of the Korean aspect construction has been explored by 

several scholars through various lenses. One study by Lee and Kim (2007) was an empirical 

approach to the -ko iss action in progress and a/eo iss continuous state construction, analyzed 

through the lens of the Aspect Hypothesis. When studying constructions related to temporality, 

22 

 
for example, imperfective aspect or continuous aspect constructions, one lens linguists have 

relied on is the Aspect Hypothesis (see Andersen, 1990, 1991; Andersen & Shirai, 1994; Bardovi 

Harlig & Comajoan-Colomé, 2020). To briefly summarize, the Aspect Hypothesis attempts to 

describe the acquisition order and usage of aspect constructions in L1 and L2 language. While 

results have been largely mixed, generally, we observe that in languages with progressive aspect, 

progressive marking is used first with activity verbs (e.g., verbs expressing activities that happen 

over a period of time, but where the endpoint is arbitrary as run in they ran around the park), and 

later with accomplishment (verbs where the action has a duration and a definitive endpoint, such 

as run in run a mile) and achievement (verbs expressing an event which takes place in an instant 

or a moment, such as recognize, die, or reach in the context of reach the top) verbs (Andersen & 

Shirai, 1996), based on Vendler’s (1957) four-way classification of a verb’s inherent lexical 

aspect. 

Lee and Kim (2007) thus explored the acquisition of the progressive continuous 

(imperfect aspect) constructions -ko iss (action in progress) and -a/eo iss (continuous state) using 

cross-sectional data from over 100 learners of Korean. Data were collected through sentence 

interpretation and guided picture description tasks. The results from their findings confirmed, 

largely as expected, that among the continuous aspect constructions the action in progress -ko iss 

develops before resultative -ko iss and -a/eo iss constructions, which is in-line with the Aspect 

Hypothesis. Further, learners exhibited less frequent usage of the continuous state -a/eo iss 

constructions than the progressive -ko iss. Given typological differences between Korean and 

English such results may be expected. Further analysis using corpora can help glean which verbs 

and verb-types may be more or less associated with each aspect construction in both L1 and L2 

23 

 
Korean. As -ko iss has a much higher rate of usage in learner language, it was selected as the 

main focus for extraction from the corpus data in the present dissertation. 

As mentioned previously, a major source of input for learners comes from the textbooks 

and materials used in class. There have been a few studies to date on the progressive in Korean 

in textbooks, and so far the general trend appears to be that continuous aspect constructions are 

not appearing in the abundance or variety that they have the potential for in L1 speech. For 

example, the resultative use of the -ko iss construction is not featured in all textbooks (Brown & 

Yeon, 2010), but when it was, it was often used without explanation and in the context of wear 

verbs regarding clothing, as the -ko iss construction can be used to express both (i) the act of 

putting on clothes and also (ii) the act of currently wearing the clothes. In some literature, this 

“resultative” meaning of -ko iss is discussed as a separate construction (e.g., Chae, 2018). In the 

present study, I am exploring the form-function of -ko iss but will only address this distinction if 

distinctive collexemes including verbs with resultative (such as to wear verbs) meanings are 

identified.  

From a usage-based perspective, including wear verbs with the -ko iss construction is 

important as it is one common use and the form significantly differs from English. However, it 

leaves teachers wanting when other instances of the resultative meaning are not included. In that 

vein, a study by Jang (2005) tallied the number of textbooks which included a discussion of the 

resultative -ko iss and found that only a fraction of the textbooks introduced the various 

meanings and semantic uses of -ko iss, and most books only touched on the standard progressive 

form across textbooks prepared for general learners of Korean and learners with specific L1s 

(e.g., Japanese). Kim (2014) carried out a comprehensive study on the change of the usage of -ko 

iss historically and diachronically, as well as using the spoken section of the Sejong Corpus to 

24 

 
identify frequency of the progressive in the corpus based on the Vendlerian (1957) categories 

(accomplishment, activity, achievement, state). Through the analysis of the distributions of the 

types of -ko iss, they found that, as perhaps expected given the Korean -ko iss construction’s 

wide variety of usage cases (e.g., action in progress, iterative progressive, narrative present, 

stative progressive, resultative, habitual, etc.), that the distribution between -ko iss functioning 

semantically as a prototypical progressive (e.g., action in progressive) or otherwise (such as 

stative or resultative meanings) were quite similar (around 40% and 45%, respectively) (p. 48). 

Thus, Kim states that due to the fact that the progressive -ko iss construction actually conveys 

not only progressive meanings, but also stative and resultative meanings, that the construction 

itself may need to be reassessed as just expressing “the general imperfective, encompassing the 

habitual and the non-Progressive use(es)…” (p. 49).  

I agree with Kim’s assessment of the Korean -ko iss construction and argue that due to its 

complex semantics, a robust analysis of what factors predict the choice of a progressive is 

necessary for not only describing the Korean language but also informing the development of 

textbooks. As mentioned, the way the progressive is introduced in textbooks is often lacking or 

incomplete, with some texts including only the purely progressive usage, others including some 

variety but without clear explanation as the function of the construction with various semantic 

meanings. In this study, I hope to build on the existing literature by marrying the findings from 

studies on the Korean imperfective and continuous aspect constructions outlined above and build 

on them using approaches that are becoming more common in the field of corpus linguistics. In 

the methodology section below, I outline the choice of corpora, predictor variables, statistical 

tools, and collostructional analysis.  

25 

 
 
III. METHOD 

In this section I outline the methodology for this study. First, details about the L1 and L2 corpora 

are provided, and a description of the textbook corpora compiled for the study is also provided. 

Then, the data extraction methods used to identify progressive and non-progressive forms for 

analyzing variation patterns within the corpora is discussed. The factors and levels for annotation 

of the data are also discussed in detail, for example, aspect (progressive versus non-progressive), 

variety (L1 Korean, L2 Korean, Textbook language), semantic domain (based on Biber et al., 

1999, Biber et al. 2021) and Aktionsart (following Deshors & Rautionaho, 2018; Rautionaho, 

2020; who borrowed Vendler’s 1957 model of Aktionsart). Finally, data analysis methods are 

discussed, including collostructional analysis (distinctive collexeme analysis), and regression 

modeling.  

3.1 Corpora description 

3.1.1 Sejong corpus 

The Sejong Corpus was a corpus of L1 Korean written and spoken langued made publicly 

available by the National Institute for Korean Language in South Korea1. The Sejong Corpus 

provides L1 Korean language corpora in both spoken and written formats (and has recently 

expanded to include other modalities such as a text message corpus). As outlined by Lee (2022) 

in the Routledge Handbook of Korean as a Second Language, the development of the corpus was 

funded by the Korean government over the span of roughly ten years (1998 to 2007). The total 

size of the corpus is about 200,320,000 ejeols2. I use the Written section of the corpus in this 

study, which in total is about 36,879,143 ejoels. The contents of this include news articles, books 

1 I used part of the Sejong Corpus for this study, however, this corpus is no longer available due to copyright issues. 
An updated corpus is now available, titled Modu-eui Malmoongchi (Korean: 모두의 말뭉치; English: Everyone’s 
Corpus) and is available online at the following URL: https://kli.korean.go.kr/  
2 An ejeol is a word and any grammatical suffixes attached in Korean. 

26 

 
 
and novels, and essays on a wide range of topics. However, due to the size of the raw corpus and 

needing to manually convert UTF-16 files to UTF-8 to make the files readable by R and 

AntConc, I randomly selected 100 files from this written corpus for analysis in this dissertation. 

Data were annotated using UDPipe (Straka et al., 2017) on a personal computer to ensure that the 

POS-tagger applied to all data is the same. The collection of the 100 files is coined as the 

KOR100. The KOR100 is 4,784,997 ejeols in size. 

3.1.2 Learner corpus 

The National Institute of Korean Language Korean Learner Corpus was selected as the learner 

corpus for the present study. The corpus is compiled by the National Institute of Korean 

Language (NIKL) and is freely available to download from the NIKL website 

(https://kcorpus.korean.go.kr). The Korean learner corpus is a large corpus which also includes 

error annotation. The corpus includes data from learners from over 100 countries and 90 

different L1 backgrounds (Lee, 2022) and is roughly 3.78 million ejeol in size. Data samples 

were provided by learners at university-level language institutions in Korea, Korean immigrant 

educational institutions, as well as universities and King Sejong institutes outside of Korea over 

from 2015 to 2021. The data were collected through collaboration at these language learning 

institutions, and an Excel file provided by the NIKL gives and overview of sample topics 

learners were prompted with when writing their essays. Students were tasked with writing about 

various prompts, and topics ranged from writing about one’s daily schedule to describing 

wedding customs, writing about their future in 10 years, and the need to install CCTV cameras in 

daycares, among others (full list available from the NIKL 2015~2021 Learner Corpus Sampling 

Information Spreadsheet). The corpus consists of both spoken and written data, though only the 

written subsections for the L1 English L2 Korean and L1 Japanese L2 Korean groups are 

27 

 
included for analysis here. Version 4.1 (released in 2021) of the learner corpus was analyzed for 

this dissertation. In total, 1,639 essays (184,181 ejeols) written by L1 speakers of English were 

included for this analysis, and 4,090 essays (495,391 ejeols) written by L1 speakers of Japanese 

were included for this analysis. 

3.1.3 Textbook corpora 

To examine the nature of language that Korean language learners are exposed to, two series of 

textbooks were selected for corpus compilation. The two textbook series selected for the present 

study are New Sogang Korean published by Sogang University Press, the recently updated 

edition of KLEAR Integrated Korean published by University of Hawaii Press. Each textbook 

was selected as they are currently used in Korean language programs in South Korea and the 

North American higher education context and an analysis of the texts within each set of learning 

materials can provide a view of the input learners are exposed to in Korean classes.  

Table 3.1 shows the number of tokens present in the textbook data stratified by level. 

Token number was identified using AntConc (Anthony, 2023). Files were manually and semi-

automatically converted into machine readable formats. Files were then converted to UTF-8 to 

allow for token counting and extraction. The textbook analysis is a frequency analysis using raw 

and relative frequencies to quantify usage of lemmas across textbook volumes. 

28 

 
 
 
 
 
 
 
Table 3.1  

Summary of data in New Sogang Series and KLEAR Integrated Korean textbook series (number 

of tokens) 

Level 1 

Level 2 

Level 3 

Level 4 

New Sogang Korean 

KLEAR Integrated 
Korean 

5802 

7386 

8730 

13679 

25519 

13704 

18836 

24298 

Total 

53730 

64224 

3.2. Part of speech tagging and extraction 

The data used in this study come from two main sources: The National Institute of Korean 

Language (NIKL) corpora, and a textbook corpus. To identify and extract all instances of the 

target constructions the data was Part-of-Speech (POS) tagged. Tagging of corpus data is a key 

step as it allows for the extraction of both the progressive and non-progressive forms of target 

verbs. While some parts of the L1 Corpus include POS annotation, running the raw data through 

POS-tagging myself has some advantages, namely that all the corpora can be POS-tagged using 

the same POS-tagging models. Following Jung (2022), I used UDPipe (Straka et al., 2016), a 

package available for R which allows for POS-tagging, tokenization, lemmatization, among 

other Natural Language Processing tasks. However, the main justification for the use of UDPipe 

at data cleaning, preparation, and extraction steps is that it includes pre-trained models of Korean 

that can be used during the POS-tagging and annotation process. I used the function 

‘udpipe_annotate’ and called for the pre-trained model for Korean (korean-gsd-ud-2.5-

191206.udpipe) to tag the data. 

After tagging the data in R using UDPipe, I used the freeware tool AntConc version 4.2.1 

(Anthony, 2023) to extract progressives and non-progressives. This was a multi-step process. 

First, I extracted the examples of the progressive -ko iss and compiled a list of lemmas appearing 

29 

 
 
 
in the progressive (for this study, I focus on present progressive for the collostructional analysis 

and regression modeling, so those were extracted). Then, following previous corpus studies on 

the progressive, I extracted corresponding non-progressive examples in the simple present based 

on the list of lemmas that co-occur with -ko iss. As Korean is agglutinative, regular expressions 

had to be written to be able to call for verbs which had a grammatical morpheme both attached 

and unattached. As an example, the verb to party in Korean is patihada (파티하다). Pati 

corresponds to party, and hada corresponds to the English equivalent of do (so the singular verb 

means to party). However, sometimes, the choice to include a grammatical morpheme on the 

lexical part of the verb (here, pati) is also possible, resulting in the form patireul hada, with the 

accusative case marker reul 를attached directly to the lexical word pati, and causing a space 

between the two parts of the verb (파티하다→파티를하다). This is a unique feature of hada 

verbs (verbs which include hada 하다), and so, to ensure accurate extraction, two versions of the 

regular expressions were submitted to AntConc to call for the target verb forms in the POS-

tagged data. 

To illustrate data extraction, I will borrow an example from Rautionaho (2020) who 

extracted key verbs appearing in stative progressive constructions. Take the verb want as an 

example. Rautionaho illustrates how a regular express used in POS tagged data can extract first 

the forms in the non-progressive using the following expression first: 

\bwant\S*(VBD|VB|VBZ|VBP|VBN)\b (p. 188), as the expression includes tags for verb forms 

such as present or past tense. The second step, then, is to swap the tags for the present participle 

form (VBG) which aids in identifying instances of the progressive. These extractions can then be 

organized in a spreadsheet for further cleaning and annotation. Of note is that Rautionaho 

30 

 
employs a two-step process in her extraction method as it allows for keeping data organized 

(namely, keeping the target constructions and the rest of the data separate from each other). In 

the present study, I also employ a two-step extraction process for each target construction. 

For stative progressives in Korean, I refer to Yeon and Brown (2011) who identify certain verbs 

in Korean which appear with the progressive. Namely, those verbs are: know, not know, love, 

believe, want, remember, and feel (p. 215). Instances of stative progressives in L1, L2, textbook 

corpora will thus be quantitatively and qualitatively explored. It is possible that stative 

progressives not listed above may appear in the data.  

Examples  of the regular expressions used include:  

1.  \b 받 VV EF\b (to call for non-hada verbs; here, the verb stem is 받 bad, and it can be 

swapped out for another form or left blank to call for all verbs in the written simple 

present form). 

2.  \b 생각 NNG |한다| XSF EF\b (to call for intact hada verbs; here, 생각 to think is used as 

an example) 

3.  \b 생각 NNG JKO |한다| VV EF\b (to call for hada verbs with accusative case marking 

attached to the lexical part of the verb; the tag JKO calls for accusative case marking) 

4.  \bNNG JKS 된 VV EF\b (to call for the verb 되다 to become in the data; in the simple 

present, when this verb means to become, it corresponds to the nominative case marker 

which the addition of JKS calls for). 

3.3. Annotation of explanatory variables 

Annotation and coding of explanatory variables (predictors) are discussed in this section. An 

overview of all explanatory variables is listed in Table 3.2. The explanatory variables in this 

31 

 
 
study include: aktionsart category, animacy of the subject, aspect (progressive or non-

progressive), semantic domain of the verb, and variety (L1/L2 Korean).  

Table 3.2. 

Proposed predictors to annotate for the logistic regression 

Predictor 

Levels (adapted from Rautionaho et al. (2018), Rautionaho (2020). 

Aktionsart 
(Vendler, 
1957) 
Animacy 

Aspect 
(dependent 
factor) 
Semantic 
domain (Biber 
et al., 1999)  

Variety 

Accomplishment, achievement, activity, stative (e.g., Vendler, 1957) 

Animate, human, inanimate 

Progressive -ko iss,  
Non-progressive  

Activity, aspectual, causative, communication, existence, mental, occurrence 

(i) 
(ii) 

(iii) 

L1 (NIKL Sejong),  
L2 (NIKL Learner Corpus: L1 English and L1 Japanese 
subsections),  
Textbook 

Aktionsart categories are based on Vendler’s (1957) classification of verbs is a “four-way 

classification of the inherent semantics of verbs” (Andersen & Shirai, 1996, pp. 531-532) based 

on a verb’s inherent lexical aspect. which includes four distinct semantic types: states, activities, 

accomplishments, and achievements. These categories are determined by three elements, namely, 

a verb’s dynamism, durativity, and telicity and have been used for lexical verb classifications in 

numerous studies (e.g., Rautionaho & Deshors, 2020; Salaberry & Shirai, 2000). It is important 

to note that aktionsart annotation must consider the context in which the verb occurs. For 

example, run can be either an activity (e.g., she is running in the park) or an achievement (e.g., 

he ran a mile). That is to highlight that the context in which the verb occurs, and the semantics of 

the verb phrases, must be considered. 

32 

 
 
 
 
 
The aktionsart classification, then, falls to the verb’s inherent lexical aspect and 

temporality, the duration the action the verb describes takes place, and the endpoint. Telicity 

refers to whether a verb has an endpoint: telic verbs have endpoints and fall into the achievement 

or accomplishment aktionsart categories. Atelic verbs do not have a clear endpoint and are thus 

categorized as activity verbs with aktionsart categories. Breaking telic verbs down further, 

whether they are achievements or accomplishments depends on the durativity of the verb, where 

punctual verbs with abrupt endpoints are classified as achievements, and verbs where the 

endpoint takes some time to culminate are categorized as accomplishments (accomplishment 

verbs are also often said to be verbs with “goals”). States are those verbs which are durative and 

describe the state of something, such as to know. Aktionsart has been useful in corpus-based 

studies on the progressive in interlanguage and World Englishes. Generally, it has been found 

that the progressive is predicted (or “triggered”) by verbs whose subjects are animate and where 

the lexical category is activity (e.g., Biber et al. 1999). However, as the Korean language allows 

for statives to occur in the progressive there may be some variation between L1, L2, and 

textbook language that is worth investigating with aktionsart categories as a predictor. Table 3.3 

provides a list of the aktionsart categories with example sentences based on Andersen and Shirai 

(1996). Of note is how depending on the context, a verb can be categorized in different aktionsart 

categories, such as run. This demonstrates the importance of considering the entire verb phrase, 

not just the verb itself, when annotating for aktionsart.  

33 

 
 
 
 
 
Table 3.3. 

Aktionsart categories with examples 

Aktionsart category 
Accomplishment: telic, 
time span of action has a 
clear terminal endpoint. 

Verb: Read. 
Sentence: I read the magazine in an hour. 

Example 

Achievement: telic, 
endpoint is punctual, and 
the event takes place 
instantaneously at a single 
point in time. 
Activity: atelic, duration of 
a period without a terminal 
endpoint, or an endpoint 
which is arbitrary.  

State: durative, describe a 
state. Verbs lacking a 
habitual reading in simple 
present are states. 

Verb: Run. 
Sentence: Brittany ran a mile. 
Verb: Recognize. 
Sentence: I suddenly recognized his voice on the phone. 

Verb: Die. 
Sentence: She died in her home last Tuesday. 
Verb: Run. 
Sentence: He is running in the park. 

Verb: Play. 
Sentence: Boram is playing with her doll. 
Verb: Love. 
Sentence: Romeo loves Juliet. 

Verb: Want. 
Sentence: Serena wants to go back to college.  

The predictor animacy refers to the verb’s main subject and whether it is alive and 

sentient, though in linguistic research animacy falls across a spectrum rather than being binary 

animate/inanimate. The progressive construction was first explored in terms of animacy of the 

subject by Strang (1982). Strang coded animacy across a continuum, first with “subjects [that] 

which are human or otherwise viewed as capable of activity,” (p. 443): (a) human, (b) quasi-

human and/or animal, and finally (c) inanimate subjects. Other studies have included more 

factors within the animacy category, such as Zaenen et al. (2004) who discussed annotating for 

subject animacy distinctions including “collectives of humans when displaying some degree of 

group identity,” computers as “intelligent machines,” and even vehicles (pp. 3-5). For the 

purposes of the present study, animacy will be coding as human, animate, or inanimate.  

34 

 
 
 
 
 
 
 
The predictor semantic domain pertains to the semantic meaning of a verb in context. 

First discussed in 1999, and again, in 2021 by Biber, Johansson, Leech, Conrad, and Finegan, the 

seven-level classification of verbs is based on a verb’s core meanings, or “the meaning that 

speakers tend to think of first” (p. 359). It is important to consider a verb not only in isolation but 

in the context in which it appears. For example, a verb such as get in English can mean obtain, 

but it can also mean become (consider I got the money from him yesterday versus I got so scared 

when I thought he didn’t have the money). Thus, when annotating for semantic domain, verbs are 

considered in the context of the sentence in which they appear. 

Table 3.4. 

Semantic domains and descriptions based on Biber et al., 1999 and 2021 

Semantic domain category 

Descriptions 

Activity verbs 

Communication verbs 

Mental verbs 

Causative verbs 

Denote actions/events associated with a choice. 
Buy, carry, go, leave, run, work… 
Transitive/intransitive 
Special subcategory of activity verbs that involve speaking and 
writing.  
Ask, announce, call, discuss, explain, say, shout, speak, 
suggest, yell, tell, write… 
Denote activities and states experienced by humans, but do not 
involve physical action (and not always volition). Subject is 
usually the recipient. Cognitive and emotional meanings 
included. 
Think, know, love, want, see, taste, read, hear 
Indicate that the person or inanimate object brings about a new 
state of affairs.  
Allow, cause, enable, force, help, let, require, permit 

35 

 
 
 
 
 
 
 
 
Table 3.4 (cont’d). 

Occurrence verbs 

Existence verbs 

Aspectual verbs 

Also called verbs of simple occurrence in Biber et al. (1999, 
2021). Report events that occur apart from any volitional 
activity.  
Become, change, happen, develop, grow, increase, occur 
Also called as existence and relationship verbs.  
Existence: Be, seem, appear 
Relationship: contain, include, involve, represent 
Characters the stage something is at, or the progress of an event 
or activity 
Kept, stopped, started, began, continue 

3.4. Collostructional analysis 

Collostructional analysis, broadly, is a family of statistical methods which allow for the 

measurement of the degrees of attraction between words and grammatical constructions (see: 

Gries & Stefanowitsch, 2003; Stefanowitsch & Gries, 2004). The name comes from the 

combination of construction and collocation (Gries & Stefanowitsch, 2003, p. 100) as the aim of 

the method is to assess collocation patterns between words and constructions (distinctive 

collexeme analysis), or between words within constructions (co-varying collexeme analysis). 

This is useful when assessing a verb which may appear in multiple constructions with a similar 

meaning as “[the verb] may ‘alternate’ between two constructions if (or to the degree that) the 

verb’s meaning is compatible with the meanings of both constructions” (Gries & Stefanowitsch, 

2004). This is true in the case of Korean where oftentimes the simple present can be used 

interchangeably with the progressive -ko iss construction in many cases. Verbs which exhibit a 

preference for a construction based on their calculated association strengths are referred to as 

distinctive collexemes of that construction.  

Such assessments of variation have been undertaken in corpus studies on L1/L2 English. 

In order to assess the progressive versus non-progressive alternation, for example, Rautionaho 

(2020) employed distinctive collexeme analysis (DCA). Specifically, she targeted co-occurrence 

36 

 
 
 
patterns of stative verbs in the two grammatical constructions (progressive versus non-

progressive) to assess which stative verbs are attracted to the progressive construction in 

different varieties of English. To assess the collostructional strength of a words to constructions 

using DCA, the absolute frequencies (of words in the construction) are assessed alongside the 

observed and expected frequencies in each construction (Hilpert, 2006). To run a DCA, Gries 

(2022) provides an R Script, Coll.Analysis 4.0, which allows the analyst to submit the data tables 

including the words extracted from each construction. It is important to make sure that prior to 

this the data has been adequately cleaned so that all target words are lemmatized in the same way 

to accurately account for their frequencies in each construction and that all exemplars are 

included (for raw frequency counts). Once the data tables are loaded into R using the script, the 

data undergoes a Fisher-Yates test, which provides the analyst with collostructional strength 

scores. Higher collostructional strength scores correspond to stronger associations between the 

words and the constructions, and likewise suggest higher entrenchment of the syntax-lexis links 

between said words and constructions in the speaker’s mind. In this way, as collostructional 

analysis is able to account for more than just raw frequencies: 

“…it [collostructional analysis] identifies not only the expressions which are frequent in 

particular constructions’ slots; rather, it computes the degrees of association between the 

collexeme and the collostruction, determining what psychological research has become known as 

one of the strongest determinants of prototype formation, namely the cue validity of, in this case, 

a particular collexeme for a particular construction.” (Stefanowitsch & Gries, 2003, p. 237). 

In short, this means that collostructional analysis when used with L1 and L2 corpora, allows 

language researchers to assess which combinations of words and constructions are “highly 

37 

 
characteristic” (p. 237), and thus can aid in the development of teaching materials and lesson 

planning for language teachers.  

Finally, when creating tables to summarize the results of the collostructional analysis and 

provide English definitions for all distinctive collexemes, English definitions were checked by 

referring to the Naver Korean Dictionary (available online at https://dict.naver.com/) and by 

manually inspecting the data to ensure polysemous words were separated (for example, multiple 

entries of the verb form sseu are in the table due to the verb’s polysemous nature). 

3.4. Logistic Regression 

When exploring a dependent variable with two outcomes, corpus linguists can employ (binary) 

logistic regression modeling to explore what explanatory variables may influence the choice of 

construction A or construction B. In this case, I follow statistical design from previous studies 

which use logistic regression to explore what factors may influence the choice of a learner or a 

first language speaker of a language to use the progressive or the non-progressive in their 

writing. I follow guidelines for planning, preparing, and interpreting the model as they are 

presented in Brezina’s (2018) Statistics in Corpus Linguistics: A Practical Guide. The binary 

dependent variable is the choice of a progressive or a non-progressive, and the explanatory 

variables include animacy, aktionsart, semantic domain, and variety.  

This study addresses the following research questions: 

Research questions 

1.  What are the distinctive collexemes of the progressive and non-progressive in L1 and L2 

Korean?   

38 

 
 
2.  Do any explanatory variables related to the verb (aktionsart, semantic domain, animacy) 

or speaker (variety) predict the use of the progressive -ko iss construction in L1 and L2 

Korean?  

3.  What verbs are most commonly used and taught in the progressive in Korean language 

textbooks?  

39 

 
 
 
 
4.1. Statistical approach 

IV. RESULTS 

I closely follow Rautionaho et al. (2018), Deshors and Rautionaho (2018), Rautionaho (2020), 

and Jung (2022) when road mapping the statistical design for this study. Rautionaho and 

collaborators’ studies serve as a mentor text for assessing the (non)progressive alternation 

between L1/L2 varieties as they employ both collostructional analysis (in particular, distinctive 

collexeme analysis) in tandem with robust statistical methods such as regression modeling.  

4.2. Research question 1 

Addressing research question 1, I discuss the results of the distinctive collexeme analysis for L1 

and L2 Korean data. Each group (L1 written Korean, L1 English L2 written Korean, and L1 

Japanese L2 written Korean) is discussed in its corresponding section. After discussing each 

group separately, comparisons between L1 and L2 results are made as appropriate. The tables 

with the list of distinctive collexemes (verbs that exhibited a preference for either the progressive 

or the non-progressive construction) are included in the appendix at the end of this dissertation 

due to their length. The prose in the sections below describe key highlights from the 

collostructional analysis. In particular, I go into detail stating (i) which verbs had an attraction or 

preference for the progressive and the non-progressive, and (ii) explore the semantic domains the 

verbs were categorized in upon qualitative analysis of the distinctive collexemes. I then (iii) 

provide key examples from the corpora to illustrate how verbs are used in each construction, 

placing emphasis on stative and mental verbs such as al (know) in particular for their potential 

usefulness in Korean language teaching and materials development. 

40 

 
4.2.1. Analysis of L1 Corpus: Distinctive Collexemes for the (non)progressive in L1 Korean 

Written Data 

Table A-1 in the appendix shows the distinctive collexemes for the progressive on the left 

and the non-progressive on the right. In total, 256 distinctive collexemes were identified for the 

progressive, and 131 distinctive collexemes were identified for the non-progressive. The ranking 

is calculated by comparing a lemma’s observed frequency with its expected frequency in each 

construction, as well as the total number of lemmas in each construction. A collostructional 

strength of 1.3 or greater is considered significant (Hilpert, 2006), and such lemmas are called 

distinctive collexemes of the construction.  

Verbs and their preferences for the (non)progressive construction are visually displayed 

in Figure 4.1 which can be interpreted in the following way: The x-axis labeled logged co-

occurrence frequency exhibits frequency of the lemma, and the farther to the right a lemma falls 

indicates its higher frequency. The y-axis labeled association (log odds ratio) is a visual 

representation of a lemma’s preference for the (non)progressive. To interpret a lemma’s 

preference for either construction on the figure, start from the dashed line in the middle (0 on the 

y-axis). Lemmas appearing above the dashed line were attracted to the progressive, and lemmas 

appearing below the dashed line were attracted to the non-progressive. As an example, take the 

verb moreu (to not know), which falls towards the bottom right of the figure. Moreu has a 

preference for the non-progressive construction (it falls below the dashed line), with a 

coll.strength score of 727. As moreu has a high coll.strength score and preference for the non-

progressive, it is a distinctive collexeme of the non-progressive (i.e., the verb moreu is attracted 

to the non-progressive. Likewise, looking above the dashed line the stative verb gaj (have) is 

clearly visible. Falling above the dashed line indicates its preference for the progressive 

41 

 
 
construction (coll.strength of 39.21). Using Table A-1 (located in the appendix, tables include 

English translations of distinctive collexemes) and Figure 4.1 in tandem provides a 

comprehensive view of the verbs appearing across the progressive and non-progressive 

constructions and their preferences for either construction in L1 Korean written corpus data can 

be ascertained. Of note is that there were more distinctive collexemes found for the progressive 

than the non-progressive.   

Figure 4.1. 

Visual representation of lemmas, their frequencies, and preference for the (non)progressive

Exploring the verbs attracted to the progressive and non-progressive in L1 Korean 

writing, we see a variety of verbs which fall into various semantic domains (based on Biber et 

al., 1991 and 2021) well represented in both constructions.  

Starting with the non-progressive construction, verbs fell into the following semantic 

domains:  

•  activity verbs (e.g., deuleoo – come in; manna – meet),  

42 

 
 
 
•  communication verbs (e.g., malha – speak; haeseogdwe – be interpreted; 

seolmeyongha – explain; gangjoha – emphasize; seoneonha – declare),  

•  mental verbs (e.g., moreu – to not know; bara – hope; johaha – like; jeulgi – 

enjoy, sarangha – love; weonha – want),  

•  causative verbs (e.g., heoyongha – permit),  

•  occurrence verbs (e.g., na – happen; pyeolcyeji – spread; dalha – reach (e.g., a 

level of something), dwe – become), 

•  aspectual verbs (e.g., sijagha – start; ggeutna – end).  

Of note is that in the current analysis, the existence/relationship semantic domain (e.g., 

represent, include, or contain) did not yield any distinctive collexemes in the L1 written data.  

Moving on to verbs that were attracted to the progressive, a variety of semantic domains 

are also well represented by verbs appearing in the progressive. Distinctive collexemes fell into 

the following semantic domains:  

•  activity verbs (e.g., sa – buy; dalli – run; moeu – gather; pal – sell; mojibha – recruit),  

•  communication verbs (e.g., jeonha – to tell/convey or pass on information; dabbyeonha – 

reply; nonha – discuss),  

•  mental verbs (e.g., bo – see; al – know; neuggi – feel; nuri – enjoy; uryeoha – be 

concerned or fearful; insigha – be aware; gominha – worry; mid – believe),  

•  causative verbs (e.g., chujinha – push ahead with or promote something to happen),  

•  occurrence verbs (e.g., byeonhwaha and baggu – change; jeunggaha – increase),  

•  existence verbs (e.g., daebyeonha – represent; mangraha – include or contain), 

•  aspectual verbs (e.g., beoli – start/begin; geuchi – stop; gyesogdwe – be continued; 

gyesogha – continue).  

43 

 
 
Qualitatively comparing the distinctive collexemes found in the progressive and non-

progressive in the L1 Korean written data shows that each semantic domain has unique verbs 

associated with them in each construction. For example, activity verbs in the non-progressive are 

largely verbs which can happen in a moment, for example come in or meet, whereas verbs found 

to be distinctive collexemes of the progressive inherently allow for a longer period of time, such 

as gather/collect and recruit. Notably, the verb come in (deuleoo) in the non-progressive was 

often used in the phrase it comes into my eye (눈에확들어온다), which can be translated 

idiomatically as it catches my eye in English. 

(5)  

4BH0004.txt 

Korean: 그런데펼쳐진일기장의왼쪽페이지가갑자기내눈에확들어온다.

English: However, in the open diary, the left page suddenly caught my eye/attention 

(literally: entered in my eye).  

The activity verbs found in the progressive were, as expected used to express an action 

occurring over a larger period of time as opposed to a moment, and with an inanimate subject 

(showing variety in animacy of subject): 

(6) 6BA02D33.txt 

Korean: 지구기온상승과기상이변을일으키는온실가스인이산화탄소농도의국내

증가속도가일본,중국등주변국가를크게앞지르고있는것으로나타나비상한

관심을모으고있다. 

English: The rate of increase in the concentration of carbon dioxide, a greenhouse gas 

that causes rises in temperatures and extreme weather events, is far exceeding that of 

44 

 
 
neighboring countries such as Japan and China, and thus is gathering/drawing extreme 

attention.  

Distinctive collexemes in the communication verb category in the (non)progressive also 

exhibited unique trends in their usage. First, in the case of the non-progressive, the verbs 

appearing in the communication semantic domain were largely based around disseminating 

information (e.g., haeseogdwe – interpret, seolmyeongha – explain, seoneonha – declare). 

Notable is that these verbs imply a one-way transfer of information from the speaker to the 

listener(s): 

(7) 4BJ01001.txt 

Korean: 지은이는우리가당연하게여기면서살아온근대자본주의세계자체,그리고

그것을지탱해온자유주의라는거대한이데올로기,그리고이에맞서온저항의지배적

형태모두에심각한위기가발상하여더이상그생명을지속하기어려워졌다고

선언한다.

English: The author declares that a serious crisis has arisen in both the modern capitalist 

world itself, the great ideology of liberalism that has sustained it, and the dominant form 

of resistance against it, making it difficult to sustain its life any longer. 

As can be seen in the above example, much of the usage of communication verbs in the non-

progressive tended towards conveying information, without necessarily requiring an interaction 

or reaction from the intended listener(s). On the other hand, in the progressive, communication 

verbs were largely interactional and used to describe exchanges between parties, passing 

information along, debating, and giving responses. As an example, take (8) which shows the verb 

45 

 
 
jeonha (to pass along or convey information) being used with the -ko iss construction to express 

conveying new facts and ideas by the author on various artistic mediums.  

(8) 4BJ01001.txt 

Korean: 또하나,저자가갖고있는건축을비롯한미술,사진,음악,오페라에대한

저자의식견은풍부한교양을제공해줄뿐만아니라,새로운흥미로운사실들을

전하고있다.

English: In addition, the author's insight into art, photography, music, and opera, 

including architecture, not only provides a rich culture, but also conveys new interesting 

facts. 

Mental verbs, or those verbs describing activities or states experienced by humans, 

follows with several distinctive collexemes in the non-progressive. In this category, the pair of 

verbs al (to know) and moreu (to not know), both stative verbs, were found to have a preference 

for different constructions. Al (to know), was largely associated with the progressive 

(coll.strength of 64.84), and moreu (to not know) was associated with the non-progressive 

(coll.strength of 727 – moreu was also the distinctive collexeme with the highest coll.strength 

score in the non-progressive distinctive collexeme list). 

(9) 2BA90A35.txt 

Korean: 그녀가무슨말을하고싶은지다알고있다. 

English: Everybody knows what she wants to say. 

*al (know) marked with progressive -ko iss 

46 

 
 
 
 
(10) 

5BA01B07.txt 

Korean: 많은사람들이아직에이즈를동성연애자나극소수문제있는

사람들의병으로만알고있다.

English: Many people still only know of AIDS as a disease affecting gay people 

or a very small number of people.  

*al (know) marked with progressive -ko iss 

(11) 

4BH0004.txt 

Korean: 앞으로는게임이나애니메이션같은멀티미디어쪽예술에서보다

중요한예술적성과가나올지모른다.

English: Going forward, it is not known if more artistic achievements may come 

about in multimedia fields such as gaming or animation.  

As can be seen in the examples, al (to know) is widely used with the progressive in the L1 

Korean written corpus, and moreu (to not know) is used at a high rate with the non-progressive. 

This distinction is notable as both verbs have been said to be stative verbs which can be used in 

the progressive -ko iss construction in Korean. However, according to the collostructional 

analysis, I find that there is a clear preference for al to be used with the progressive -ko iss, and 

for moreu to be used in the non-progressive, at least in written data.  

Stative verbs beyond al (know) and moreu (not know) were present in the data. Other 

stative verbs that were distinctive collexemes for the progressive included neuggi (feel), insigha 

(be aware), gominha (worry), and mid (believe). Given the fact that Korean allows for stative 

progressives at a higher rate than other languages, having so few stative progressives appear in 

the mental verb category was surprising, though it must be said that these are only those stative 

47 

 
 
progressives which were attracted to the progressive. It is possible that other stative progressives 

appeared at lower frequencies and were therefore not found to be distinctive collexemes. 

Nonetheless, the stative progressives found in the L1 written data help illuminate how the 

progressive -ko iss can combine with stative verbs in Korean. 

(12) 

2BA93A22.txt 

Korean:나는비행기보다는철도여행을좋아하므로매번불편을느끼고있다.

English: I prefer trips by train over plane, so I feel uncomfortable every time. 

*neuggi (feel) marked with progressive -ko iss 

Example (12) illuminates how the progressive -ko iss may be used with a mental stative verb 

when the experience continues over a period of time, such as feeling uncomfortable each time 

one does a certain activity. This is also realized with the verb insigha (to be aware, perceive, 

recognize), which is used with the progressive -ko iss when a speaker is discussing a point which 

they are aware of and recognize as important. 

(13) 

5BA01B10.txt 

Korean: 나는인간생명의존엄성에대한윤리적철학적배경이배제된

생명공학이인류의재앙이될수있다는점을깊이인식하고있다.

English: I am deeply aware that biotechnology, which excludes ethical and 

philosophical backgrounds on the dignity of human life, can become a disaster for 

humanity.  

*insigha (aware) marked with progressive -ko iss 

48 

 
 
Beliefs can also be expressed by combining the verb mid (believe) with the progressive in 

Korean. It can also be used to express a belief one holds about an event they presume will 

happen at a future time. 

(14) 

5BA01A09.txt 

Korean: 하지만LG선수들은김태환감독이그에대한대비책을내놓을것으로

믿고있다.  

English: However, LG athletes believe that director Taehwan Kim will come up 

with a countermeasure. 

*mid (believe) marked with progressive -ko iss 

Finally, when considering the mental verbs, and in particular the aforementioned stative 

verbs which are distinctive collexemes of the (non)progressive, an unexpected trend can be 

observed in regard to the supposed emotional sentiment expressed with the verbs in each 

construction. The collostructional analysis revealed that mental verbs which are distinctive 

collexemes of the non-progressive are largely associated with positive mental experiences or 

emotions: bara (hope), johaha (like), jeulgi (enjoy), sarangha (love), weonha (want/desire), are 

all verbs which can be categorized as largely positive emotional or mental experiences.  

However, zooming in on the distinctive collexemes for the progressive reveals a different 

pattern: uryeoha (be concerned, fearful), and gominha (worry) are mental verbs which are 

associated with negative emotions. While beyond the scope of this dissertation, it is an 

interesting finding that semantic meanings associated with the (non)progressive constructions 

seem to follow different trends in L1 Korean writing. 

For causative verbs, in the non-progressive the verb heoyongha (permit) appeared as a 

distinctive collexeme. In the progressive, chujinha (to promote something to happen) is the 

49 

 
 
 
 
distinctive collexeme identified. Chujinhada was often used with objects such as plan (계획을

추진하고있다, pushing ahead/promoting a plan) and other similar objects. 

(15) 

5BA01B06.txt 

Korean: 정부도의료전달체계확립을위해가벼운질병으로3 차병원을

이용하는환자에게는무거운진료비를물게하는방안을추진하고있다.

English: To establish a medical delivery system, the government is also pushing 

for/promoting a plan to impose heavy medical expenses on patients who use 

tertiary hospitals due to treat mild diseases. 

Several verbs falling into the occurrence verb category were distinctive collexemes. 

Starting with the non-progressive, one of the most common verbs was na which, while difficult 

to translate into English, is a verb which means to happen/come up, and in some cases break out 

or occur depending on the context. Unique to Korean, this verb is usually preceded by a noun 

marked with a nominative case to denote what is happening or occurring. Qualitative analysis of 

the data containing na shows the verb appearing with a wide array of nouns, including dust, 

disease, problem, and smell, among others. In each case, na is used to denote the occurrence of 

the noun. While it is beyond the focus of the present dissertation, a follow-up study could 

explore na and the lexical items it associates with through co-varying collexeme analysis 

(another type of collostructional analysis) to identify typical usage cases of na. 

Another verb which follows a similar pattern in the data is dwe (become). In fact, due to 

the form dwe appearing in a multitude of grammatical constructions in the Korean language, to 

extract only those occurrences of dwe expressing the become meaning, a separate regular 

expression had to be written to ensure that dwe was being extracted alongside nouns with 

50 

 
 
nominative case, followed by manual inspection of the extractions. Qualitative inspection of the 

data reveals dwe appearing with a multitude of nouns marked with nominative case, including 

reason (...의미가된다 – something becomes the reason for…; …도움이된다 – something 

is/becomes helpful). 

(16) 

5BA01B07.txt 

Korean: 사춘기에다리안쪽이나등에생기는튼살은조기에치료하면회복에

도움이된다.  

English: Treatment of stretch marks during puberty on the inner or back of the leg 

can be helpful.  

Finally, distinctive collexemes were identified in the non-progressive for the aspectual 

verb category, which includes verbs which denote the time or status something is occurring at 

(such as start, end, continue, etc.). In the L1 Korean written data, sijagha (start) and ggeutna 

(end) were distinctive collexemes of the non-progressive. 

(17) 

4BJ01001.txt 

Korean:대부분의인간적삶이란,자신이스스로의삶을거리를두고볼수있는

여유와창조적해결을위한판타지를갖지못해종종영원한미궁으로흘러

들어가거나비극으로끝난다.

English: As for most human lives, they are unable to put space between 

themselves and their lives to have the freedom to come up with creative solutions, 

and sometimes this causes them to fall into an endless labyrinth, which ends in 

tragedy. 

51 

 
 
 
Notably, there are no other aspectual verbs in the non-progressive apart from those which 

indicate the start or end of an event. In contrast, the progressive had start/begin (beoli), stop 

(geuchi), gyesogha (continue), and gyesogdwe (be continued) as distinctive collexemes. Perhaps 

unsurprisingly, the progressive -ko iss construction is used to express events as they are in the 

process of starting/stopping, as well as to describe their status in continuation. 

(18) 

5BA01A02.txt 

Korean: 일부 시위대들이 정유소 봉쇄를 풀었으나, 여전히 대다수는 

과도한 유류세 인하를 요구하며 연일 시위를 계속하고 있다.

English: Some protestors have lifted the oil refinery blockade, however others, 

demanding excessive oil tax cuts, are continuing to protest. 

Interim summary: In this section, I have shared the results of the distinctive collexeme analysis 

for the L1 Korean written data. I provided a table which allows the reader to view which verbs 

were exhibited a preference for either the progressive or the non-progressive. Discussing the 

results in prose, I outlined verbs which appeared as distinctive collexemes in the progressive and 

the non-progressive and discussed them in terms of semantic domains verbs are categorized 

while providing examples of verbs in context (taken from the corpus). As this section covers L1 

written data, there are more distinctive collexemes than there are for the learner data to follow. 

While it is natural to expect L1 data to include a wider variety of verbs, it must be noted that the 

amount of data is also substantially larger in the L1 corpus. In what follows, I will discuss the 

results of the distinctive collexeme analysis for the learner data, providing the list of the 

distinctive collexemes for the progressive and the non-progressive.  

52 

 
4.2.2. Analysis of L2 Corpus: Distinctive Collexemes for the (non)progressive in L1 English 

L2 Korean Writing 

Table A-2 in the appendix shows the distinctive collexemes for the progressive on the left 

and the non-progressive on the right for L1 English L2 Korean speakers. In total, there were 

twenty-three (23) distinctive collexemes attracted to the progressive -ko iss construction, and 

nine (9) distinctive collexemes showing a preference for the non-progressive.  

This can also be seen in Figure 4.2, which visually shows the relationship between each 

lemma, its frequency, and its preference for the (non)progressive. Figure 4.2 can be interpreted in 

the following way: The x-axis labeled logged co-occurrence frequency exhibits frequency of the 

lemma, and the farther to the right a lemma falls denotes its higher frequency. The y-axis labeled 

association (log odds ratio) is a visual representation of a lemma’s preference for the 

(non)progressive. To interpret a lemmas preference for either construction on the figure, start 

from the dashed line in the middle (0 on the y-axis). Lemmas appearing above the dashed line 

are associated with the progressive, and lemmas appearing below the dashed line correspond 

with the non-progressive. Thus, as a simple example, looking to the very bottom right of the 

figure, the lemma saenggagha (to think) is the most frequent among lemmas appearing in the 

non-progressive as it is farthest to the right, and it exhibits a strong preference for the non-

progressive, as it is far below the dashed line.  

53 

 
 
 
 
 
 
Figure 4.2.  

Visual representation of lemmas, their frequencies, and preference for the (non)progressive 

Exploring the verbs which appeared in the progressive in the L1 English L2 Korean 

learner data reveals a variety of semantic domains (based on Biber 1999 and 2021) being 

represented. In the progressive, semantic domains include: 

•  occurrence verbs (e.g., manhaji and jeunggaha – increase; byeonha – change),  

•  mental verbs (e.g., al – know, gominha – worry, neuggi – feel),  

•  activity verbs (e.g., ilha – work).  

There was a lack of the following semantic domains in the progressive data: communication 

verbs, causative verbs, existence verbs, and aspectual verbs.  

Turning to the verbs in the non-progressive, the only semantic domains covered are: 

•  activity verbs (e.g., meog – eat; ju – give),  

•  mental verbs (e.g., saenggagha – think; bo – see; deud – listen).  

54 

 
 
 
As the distinctive collexeme analysis allows an analyst to identify verbs which have a 

high association strength and preference for a particular construction, this suggests that when 

verbs appear in both constructions, their usage in the learner data will tend towards the 

progressive as opposed to the non-progressive. Further, it seems that in the case of L1 English L2 

Korean writers, the associations between certain semantic domains and a variety of usages are 

stronger with verbs in the progressive than the non-progressive.  

Digging deeper into the verbs associated with the (non)progressive in the L1 English L2 

Korean learner data, unsurprisingly, many of the verbs are used to indicate a change occurring 

over time. For example, the verb most highly associated with the progressive in the learner data 

is manhaji (increase, grow), which was often used to express changes on a societal level: 

(19) 

sample_9841.txt 

Korean: 하지만 인터넷을 무조건 믿는 사람이 많아서 인터넷 발전으로 

인해 사람들이 사기꾼 같은 사람들은 신뢰하는 경우가 많아지고 있다.  

English translation: However, as there are many people who just believe in the 

internet, through the development of the internet the amount of people who 

believe scammers is increasing. 

Further qualitative analysis also revealed some errors in learners’ usage of the verb manhaji with 

the progressive. An intransitive, the verb manhaji describes an increase or growth, and thus the 

argument of the verb should take the nominative case marker in Korean. However, some learners 

in the L1 English category exhibited errors in their usage of grammatical markers with verbs in 

the progressive: 

55 

 
 
(20) 

sample_30998.txt 

Korean: 명품소비자들은대부분여자있었는데요즘남자들도명품에대한

관심을많아지고있다.

English translation: Consumers of brand-name/designer products were mostly 

women, but these days, interest in brand-name products by men is also 

increasing. 

Interestingly, the lemma directly following manhaji came out to be jeunggaha, which 

also means increase. However, its overall lower association strength and frequency in the learner 

data could be due to it being taught at a more advanced level, and thus learners may have been 

exposed to the word less.  

Continuing in the investigation of distinctive collexemes in the learner data reveals an 

interesting trend with mental and stative verbs exhibiting a preference for and a strong 

association strength with the progressive construction in Korean. Several mental and some 

stative verbs, including gominha (worry), al (know), neuggi (feel), and a physical stative verb sal 

(live) exhibited a preference for the progressive. This is an interesting finding as Korean is 

known to allow for stative verbs at frequencies higher than English, so learners exhibiting usage 

of stative progressives in their Korean writing is a positive sign for acquisition of the form-

meaning association between states and the progressive construction in learner Korean.  

(21) 

sample_29420.txt 

Korean: 현대부모님들도외국어학습을시작할까고민하고있다.

English: Parents in modern times are worried (worrying) about starting foreign 

language acquisition. 

56 

 
 
 
 
(22) 

sample_31034.txt 

Korean: 내가중국하고영국에서도살아*본적이있어서서양문화와아시아

문화를잘알고있다.

English: Because I have experience living in both China and England, I know 

both western and Asian cultures very well. 

Note: * denotes a correction in spelling. 

As can be seen in the examples above, L1 English learners of Korean exhibit use of stative 

progressives with both stative verbs that can take the progressive in English (e.g., it is not 

unheard of for a sentence such as parents are worrying about X in English), as well as verbs that 

typically do not co-occur with the progressive, such as al (know).  

The exploration of stative verbs appearing in the learner data is more notable when 

compared with the stative verbs that are distinctive collexemes in the non-progressive. Among 

the verbs which appeared in both constructions, the only stative verb that was found to be a 

distinctive collexeme in the non-progressive was saenggagha (think), which exhibited an 

extremely strong association strength and preference for the non-progressive that was higher than 

any other singular verb’s preference for either construction in the L1 English data.  

In terms of form-meaning mappings, results show that L1 English writers of Korean 

associate the progressive with prolonged duration of an event or action (e.g., increasing, 

preparing, changing, attending, becoming, disappearing) and mental states (e.g., worry, feel, 

know, expect). On the other hand, verbs found to be significant collexemes of the non-

progressive, with the exception of saenggagha (think), are by and large (physical) actions (e.g., 

see, eat, give, oppose, send, drink, use, listen). This distinct difference between the form-

meaning mapping of the (non)progressive in Korean suggests that learners may associate the 

57 

 
 
 
progressive form with usages of prolonged duration and mental states, and the non-progressive 

form with physical action verbs. Diving into the usage cases of such action verbs, we can see that 

learners use the non-progressive for habitual actions:  

(23) 

sample_3890.txt 

Korean: 버스나지하철을탈때헤드폰으로항상음악을듣는다.

English: When I ride the bus or subway I always listen to music with headphones. 

Further, learners are also correct in associating the verb bo (to see) in the non-progressive 

with its usage of expressing how one views a certain state or situation. In other words, for the 

verb bo (to see), learners have acquired its usage which extends beyond the simple to see/to 

watch function and are able to use it to express their views. 

(24) 

sample_6714.txt 

Korean: 사회가빠르게변화하면서사람들의관심분야도변하기때문에

전통문화를조금식현대화시키는것은모두에게좋다고본다.  

English: As society changes rapidly so too do people’s interests, and so I view the 

change in traditional culture to something more modern as a good thing.  

Interim summary: Overall, the distinctive collexeme analysis of verbs in the progressive and 

non-progressive in L1 English L2 Korean writing show interesting trends. Most notable is the 

trend of the progressive construction -ko iss being associated with stative and mental verbs when 

compared with the non-progressive form. Results suggest that L1 English learners of L2 Korean 

are able to overcome the obstacles that may be presented by typological differences between 

English and Korean, namely that Korean allows for more stative verbs than English. While it 

was hypothesized that learners from the L1 English background would demonstrate and overall 

lack of stative verbs in their writing, that fact that several stative verbs appeared in their writing 

58 

 
 
 
is a positive sign for the accurate development of the form-meaning mappings of the progressive 

construction in learner language. Perhaps most notable is the learner usage of al (to know), a 

stative verb in Korean which almost never takes the progressive in English. Overall, results 

suggest a positive trend towards felicitous usage and promising acquisition patterns of the 

progressive -ko iss construction in L1 English L2 Korean. Finally, notable in this set of learner 

essays is that when it comes to verbs which appear in both constructions, more verbs appearing 

in those constructions are found to be distinctive collexemes of the progressive as opposed to the 

non-progressive. From a usage-based perspective, it could be the case that the input learners 

receive, either through textbooks or interactions with native speakers, is that these verbs simply 

appear more with the progressive than the non-progressive, and thus when focusing on verbs that 

can appear in both, it is natural that more distinctive collexemes are found in the progressive.  

4.2.3. Analysis of L2 Corpus: Distinctive Collexemes for the (non)progressive in L1 

Japanese L2 Korean Writing 

In this section, the findings from the distinctive collexeme analysis of the L1 Japanese L2 

Korean data are discussed. 

As can be seen in Table A-3, there is an uneven distribution of the number of distinctive 

collexemes in the progressive and non-progressive. As a distinctive collexeme analysis only 

explores those verbs which appear in both constructions, it appears that when a verb can appear 

in either construction, learners tend to associate verbs with the progressive more often and use 

the verbs in the progressive at higher frequencies. This is borne out in higher collostructional 

strength scores, yielding more distinctive collexemes for the progressive than the non-

progressive. In total, there were fifty distinctive collexemes for the progressive construction and 

fourteen distinctive collexemes associated with the non-progressive construction. 

59 

 
This uneven distribution is visualized in Figure 4.3, which visually represents each verb 

and its preference for the progressive or the non-progressive. All verbs that were submitted to the 

collostructional analysis are featured in Figure 4.3, hence why there appears to be more lemmas 

listed than can be seen in Table A-3 (Table A-3 only includes those which yielded 

collostructional/association strengths of greater than 1.3; Figure 4.3 represents all lemmas which 

were extracted and submitted to the distinctive collexeme analysis). Figure 4.3 can be interpreted 

in the following way: The x-axis labeled logged co-occurrence frequency exhibits frequency of 

the lemma, and the farther to the right a lemma falls denotes its higher frequency. The y-axis 

labeled association (log odds ratio) is a visual representation of a lemma’s preference for the 

(non)progressive. To interpret a lemmas preference for either construction on the figure, start 

from the dashed line in the middle (0 on the y-axis). Lemmas appearing above the dashed line 

are associated with the progressive, and lemmas appearing below the dashed line correspond 

with the non-progressive. Thus, as a simple example, looking to the very bottom right of the 

figure, the lemma saenggagha (to think) is the most frequent lemmas as it is farthest to the right, 

and it exhibits a strong preference for the non-progressive, as it is far below the dashed line.  

60 

 
 
 
 
 
 
 
 
 
Figure 4.3. 

Visual representation of lemmas, their frequencies, and preference for the (non)progressive 

First, when exploring the distinctive collexemes in the L1 Japanese L2 Korean data, it is 

clear that there is a variety of verb types which are attracted to both the progressive and non-

progressive construction. For example, the most common verb overall was saenggagha (think), a 

mental verb which exhibits a strong preference for the non-progressive construction. In addition 

to saenggaggha, other mental verbs with a preference for the non-progressive include moreu (not 

know), boi (be visible), bo (see), and neuggi (feel).  

(25) 

sample_6334.txt 

Korean: 큰일이생기면나중에후회할지도모른다.

English: If a big issue crops up later, I don’t know if I’ll regret it. 

In the case of the verb bo (see), learners usage of this verb to express their views and the way 

they observe the world were found in the data, for example, when discussing societal issues (the 

writer in the example below was discussing gender issues in Japan): 

61 

 
 
 
 
(26) 

sample_32001.txt 

Korean: 아마도여자의의식이남자보다앞서있다고본다.

English: The way I see it, women probably have more consciousness/awareness 

than men do.  

The non-progressive also features two communication verbs, namely malha (speak) and 

iyagiha (talk), two verbs which are often interchangeable and share similar ranking and 

collostructional strength scores (ranked fourth and fifth distinctive collexemes, with coll.strength 

scores of 22.02 and 19.33, respectively). Similar ranking suggests learners may use these verbs 

interchangeably and that they have similar levels of entrenchment in the learner language. 

Activity verbs (denoting actions and events associated with someone’s choice or own volition) 

are represented among the distinctive collexemes for the non-progressive as well, including 

verbs such as ga, meog (eat), ju (give), sa (buy), and sigsaha (have a meal). Put simply, 

distinctive collexemes associated with the non-progressive construction in the L1 Japanese L2 

Korean variety include mental, communication, and activity verbs only.  

Diving into the progressive exhibits a richer array of verb types and larger variety of 

lemmas, suggesting that verbs which appear in both the (non)progressive may tend towards the 

progressive -ko iss construction in learner Korean. Additionally, a larger variety of verb types are 

found in the distinctive collexemes for the progressive, including activity verbs (e.g., ilha – 

work), mental verbs (e.g., gominha – worry), occurrence verbs (e.g., baggui – be changed; 

baldalha – develop). Therefore, it appears that in terms of overlapping verb types (based on 

Biber 1999 and 2021 semantic domains) overlap occurs for activity and mental verbs.  

62 

 
 
 
(27) 

sample_9277.txt 

Korean: 지금의직장은스트레스를많이받긴하는데즐겁게일하고있다.

English: At my current workplace I do get stressed but I am working joyfully. 

(28) 

sample_33595.txt 

Korean: 근데어학당을졸업한후에한국에서일을할지일본에서일을할지

고민하고있다. 

English: But now, I am worrying about whether to work in Korea or Japan after 

graduating from the Korean language school. 

(29) 

sample_31394.txt 

Korean: 최근세계적으로가족의형태가바뀌고있다.

English: Recently, the form of the family unit is changing globally.  

(30) 

sample_32874.txt 

Korean:한국에서는인터넷을이용한배달이나택배가많이발달하고있다.

English: Online delivery and shipping services are developing in Korea. 

Distinctive collexemes for communication verbs were only present in the non-

progressive, and occurrence verbs were only present in the progressive. Both constructions 

yielded distinctive collexemes for stative verbs, with the stative verb with the highest 

collostructional strength in the L1 Japanese L2 Korean data being gaji (have, coll.strength 

138.10), followed by gominha (worry, coll.strength 32.34), gidaeha (expect, coll.strength 24.46), 

and mid (believe, coll.strength 11.12).  

63 

 
 
 
(31) 

sample_15460.txt 

Korean:사람에따라다른생각을가지고있다.

English: Depending on the person, the thoughts they have/hold (progressive 

marked in Korean) differ. 

(32) 

sample_34551.txt 

Korean:그중에서도*교토사투리는표준어에서는찾아볼수없는독특하고

부드러운어감을가지고있다.

English: Even among them*, the Kyoto dialect has (progressive marked in 

Korean) a unique and soft sense of language that cannot be found in standard 

language. 

*그중에서도 refers to the various dialects of the Japanese language. 

(33) 

sample_6752.txt 

Korean:보통특히여성들은명품에관심이있는듯싶다.나도관심이있고갖고

싶은욕심을가지고있다.

English: Usually, and especially, women have an interest in designer products. I 

also have (progressive marked in Korean) an interest in and greediness for 

designer products. 

Apart from gaji (have), notable examples of the stative progressive in the L1 Japanese data were 

found for the distinctive collexeme mid (believe) in students’ writing on superstitions:  

64 

 
 
 
 
(34) 

sample_13364.txt 

Korean:그것은찻울기가서면집에기웅도선다고생각해서근웅이

좋아지거나좋은일이일어난다고믿고있다.

English: If the tea leaf stands, it is believed (progressive marked in Korean) that 

good fortune will come.  

In some instances, learners used the progressive with mid (believe) to express beliefs they 

currently hold: 

(35) 

sample_35879.txt 

Korean:친구,가족,아는사람들을모든사람들을사랑하는노력을서로할수

있으면다들행복하게살수있다고믿고있다.

English: I believe (marked with progressive in Korean) if people who know each 

other put in an effort to love each other, then everyone can live happily.  

Overall, results show promising acquisition of the progressive construction in Korean, its 

usage with various verbs of various semantic domains (not limited to physical actions), and the 

development of the usage of stative progressives in Korean.  

4.2.4. Comparisons of L1 and L2 corpora and discussion of potential interlanguage transfer 

effects 

As one goal of this dissertation is to illuminate how the progressive -ko iss construction is used 

in L1 and L2 Korean, in this section, I will briefly cover some of the differences observed in the 

progressive and its usage across the L1 and L2 varieties thus far. While the focus is not 

necessarily to discuss language learning mechanisms, the results from the collostructional 

65 

 
 
analysis show which lemmas prefer -ko iss, which can be a starting point for discussing 

differences in usage based on a language user’s L1.  

In terms of interlanguage transfer effects, one of the ways to identify whether typological 

differences could be at play is to look for stative verbs appearing in the progressive -ko iss 

construction across the L1 and L2 varieties. The three varieties, L1 Korean, L1 English L2 

Korean, and L1 Japanese L2 Korean, lend themselves well to such analysis as both Korean and 

Japanese are known to allow perfective readings to be associated with the grammatical 

constructions associated with the progressive, whereas in English the progressive is never 

perfective and also describes an ongoing action or even a futurate reading in some cases (e.g., 

Lee, 2006; McLure, 1994; Yeon & Brown, 2011). Essentially, what this means is that the 

progressive constructions in Korean and Japanese can be used with stative verbs which would 

not normally be expected to take the progressive construction in a language such as English. 

Prime examples of such verbs include to know, to have, or to believe. For example, in the case of 

the verb know in Korean, it has been described as being able to take the progressive as, at one 

point, someone came to know the information, and they will remain in that state of ‘knowing’ 

that information until the moment they forget it, at which point, the state would be over (Lee, 

2006). The Japanese progressive, -te iru, functions in a similar way particularly for stative verbs, 

thus it could be anticipated that L1 Japanese learners of Korean will incorporate more stative 

progressives in their writing than their L1 English counterparts, thus providing some potential 

evidence for interlanguage transfer effects. However, given that the present study is corpus-based 

and not experimental to test for specific interlanguage transfer effects or cross-linguistic 

influences, at most I will only discuss trends that appear in the data.  

66 

 
 
To compare the results of the distinctive collexeme analysis, I first normalized the 

coll.strength scores for each variety and visualized them using bar charts. Doing so makes 

identifying trends in lemma associations with the progressive across varieties easier and allows 

for an analyst to quickly identify stative progressives appearing in each variety.  

Visual inspection of the normalized coll.strengths in Figures 4.4, 4.5, and 4.6, for all 

varieties shows a sharp drop in coll.strengths after the first two or three distinctive collexemes, 

showing that certain verbs may be more prototypically associated with the progressive -ko iss 

construction. Unsurprisingly, the L1 variety yielded the highest number of distinctive collexemes 

(256), followed by the L1 Japanese group (78) and the L1 English group (43).  

While typological similarities could be one explanation as to why the L1 Japanese data 

including more distinctive collexemes than the L1 English data, it should also be noted that these 

differences could be due to the different number of learner essays available in each corpus, so 

this trend should be considered with caution. 

A well attested stative progressive to appear in both Korean and Japanese is know (al in 

Korean; shiru in Japanese), and that verb appears in both the L1 and learner corpora. In the L1 

data, al is the ranked 18th distinctive collexeme with a coll.strength score of 64.84, confirming 

the form-function mapping of al with the progressive -ko iss construction in Korean is well 

attested in the reference corpus. In the L1 English L2 Korean corpus, al also appears, ranked as 

the 29th distinctive collexeme with a coll.strength score of 3.3, a perhaps surprising finding as it 

was anticipated that L1 English speakers would likely not use the verb know in the progressive 

due to know rarely taking the progressive in English. However, more surprising is the finding 

that al was not a distinctive collexeme in the L1 Japanese learner data. From this, there are two 

points to highlight. First is that despite a relative lack of stative progressives in English (and 

67 

 
 
 
 
especially with the verb know) it does not appear to be impeding the uptake of this form-function 

mapping in L1 English learners of L2 Korean, which is a positive sign for second language 

learners. Second, while beyond the scope of this study, it could be the case that there are factors 

at play which determine the use of a stative progressive with know in Japanese that is not 

represented in the data. For example, while this study is limited to exploring written data across 

all varieties, it could be the case that certain stative progressives are more common in both 

Korean and Japanese spoken language, which could be a reason why al is not strongly associated 

with the progressive in the L1 Japanese learner data despite a similar form being well attested in 

the learners’ L1. However, that is not to say that the Japanese data did not include al (know) at 

all. In fact, while al (know) was not a distinctive collexeme in the Japanese variety, normalizing 

the data revealed that Japanese speakers actually used al more frequently with the progressive -

ko iss than both the English and L1 Korean varieties. This suggests that, in L1 Japanese learners’ 

L2 Korean writing, they are choosing to use other verbs significantly more often than an (know). 

Put simply, Japanese learners of Korean had more verbs co-occur with the progressive, and as 

such even some key verbs (including al) were not found to be distinctive collexemes of the 

construction despite an overall higher relative frequency.  

While the L1 Japanese data lacked the verb al (know) as a distinctive collexeme, other 

common stative progressives were present in the data. For example, gaji (to have/to hold) 

appeared in the L1 data ranked 16th distinctive collexeme with a coll.strength of 69.5. The same 

verb was a distinctive collexeme for the L1 Japanese data, ranked 2nd distinctive collexeme with 

a coll.strength of 138.10. Gaji was not found to be a distinctive collexeme in the L1 English 

learner data. The verb have in Korean and Japanese (gaji and motsu) are both known to be used 

with the progressive construction, so it is unsurprising that it appeared in both the L1 data and 

68 

 
 
that it was a distinctive collexeme highly attracted to the progressive construction in the L1 

Japanese learner data. A notable point about the way the verb have functions in both languages, 

and largely in Korean, is that it can be used to denote ongoing possession of not only physical 

objects but also intangible things including thoughts, feelings, backgrounds, impressions, and so 

on. In that way, Korean and Japanese are typologically similar, however, English differs in this 

regard as the verb have would rarely be used with the progressive to express similar sentiments. 

Examples of intangible possession in L1 Korean data with progressive gaji: 

(36) 

4BH0005.txt 

Korean:그들은‘훌륭한사회'(goodsociety)에대한이미지를가지고있다.

English: They hold/have an image of a “good society.”  

(37) 

4BH0004.txt 

Korean:그래서이세가지는항상서로통하는의미를가지고있다.

English: So these three things always have a common meaning.  

(38) 

4BH00013.txt 

Korean:나는시인들에게깊은존경심을가지고있다.

English: I have a deep respect for poets. 

Example of tangible possession in L1 Korean data with progressive gaji: 

(39) 

5BA01B04.txt 

Korean:핵시대에접어들면상황은더악화된다.‘우리핵무기가지고있다.’‘너

뭐줄래?’하는식이다.

English: If we enter a nuclear era/generation the situation will become much 

worse. It could be like ‘we have nuclear weapons.’ ‘What can you give us?’ 

69 

 
Examples of intangible possession in the L1 Japanese data with progressive gaji: 

(40) 

sample_26742.txt 

Korean: 나는감시카메라설치확대에대한찬성의견을가지고있다.

English: About the expansion and installation of security cameras, I have/hold an 

opinion of agreement.  

(41) 

sample_15454.txt 

Korean:나는휴일마다낮잠을자는습관을가지고있다.

English: I have/hold a habit of napping on my days off.  

*have/hold marked with progressive -ko iss 

(42) 

sample_18316.txt 

Korean:메테인가스는사실이산화탄보다약10 배의온실효과를가지고

있다.

English: In fact, methane gas has/hold approximately 10 times the greenhouse 

effects of carbon dioxide. 

As can be seen in the examples above, L1 Japanese learners of L2 Korean are able to use the 

verb gaji (have/hold) with the progressive in ways that are in-line with how L1 speakers use the 

verb, namely with intangible objects including thoughts, feelings, effects, habits, and opinions. 

From an interlanguage standpoint, the fact that this verb in the L1 Japanese L2 Korean data was 

found to be highly associated with the progressive, but not so in the L1 English L2 Korean data 

according to the distinctive collexeme analysis, suggests that there are some interlanguage 

transfer effects at play. English lacking a system of using the equivalent English verb with the 

70 

 
progressive and with tangible/intangible objects may lead to learners having difficulty with 

uptake of this form-meaning association.  

In terms of stative verbs, of particular interest are the stative verbs noted in Yeon and 

Brown (2011) as attested verbs associated with the progressive in Korean, namely know (al), not 

know (moreu), love (sarangha), believe (mid), want (weonha), and feel (neuggi). The distinctive 

collexeme analysis revealed the following trends in usage of these specific verbs with the 

progressive construction in Korean as shown in Table 4.1. Table 4.1 outlines the key verbs 

mentioned in Yeon and Brown, highlighting which construction each verb appeared in for each 

variety (the progressive or the non-progressive). If a verb appeared in and preferred either 

construction in each variety, it is marked with a plus sign ‘+’. For example, looking at the first 

row where al (know) is listed, it is clear that al (know) appeared in the L1 Korean data and that it 

exhibited a preference for the progressive -ko iss construction. Continuing on, the first major 

trend observed in the data is that all key verbs appeared in the L1 data. However, of note is that, 

among those verbs, only al (know), mid (believe), and neuggi (feel) were found to have a 

preference and association for the progressive construction in the distinctive collexeme analysis. 

Focusing on the learner data reveals some differences between L1 and L2 language. First of all, 

while English generally lacks the verb know being used with the progressive construction (e.g., it 

would be unnatural for a native speaker to say I am knowing all about that in English), the L1 

English speakers exhibited usage of the verb with the progressive construction. Perhaps 

surprisingly, this verb was not found to be a distinctive collexeme of the progressive in the L1 

Japanese learner data, and this is surprising as Japanese allows for the use of the progressive -te 

iru (Japanese progressive construction) with the verb know. For the L1 English learner data, after 

al (know), the only other verb in the list that exhibited a preference for the non-progressive was 

71 

 
 
neuggi (feel), with a preference for the progressive construction. Despite having a low number of 

distinctive collexemes from the Yeon and Brown list, the distinctive collexemes that do appear in 

the L1 English data follow similar collostructional patterns with the L1 Korean data. For the L1 

Japanese data, only two verbs from the list were associated with the progressive: sarangha (love) 

and mid (believe). Both of these verbs in Japanese are able to take the progressive construction in 

the Japanese language (e.g., aishiteiru “I am loving/I love you”, and shinjiteiru “I am 

believing”). As such phrases are possible in Japanese (though less common in English), this may 

be evidence for interlanguage transfer effects allowing for L1 Japanese speakers to acquire and 

use more stative verbs with the progressive -ko iss in their L2 Korean writing. However, notable 

is that in the L1 data, the verb sarangha is actually associated with the non-progressive. While 

this may be surprising at first, it is important to note that this dissertation only focuses on written 

language in all varieties. It is possible that such construction will appear more in spoken 

language (discussed more in limitations and future directions).  

72 

 
 
 
 
 
 
 
 
 
 
 
Table 4.1.  

Summary of key verbs appearing in the progressive -ko iss construction as distinctive collexemes 

across varieties based on Yeon and Brown (2011) 

Verb 

L1 written corpus 

L1 English L2 
Korean 

L1 Japanese L2 
Korean 

Prog 

Non-prog 

Prog 

Non-prog 

Prog 

Non-prog 

al – to know 

moreu – to not know 

sarangha – love 

mid – believe 

weonha – want 

neuggi – feel 

+ 

+ 

+ 

+ 

+ 

+ 

+ 

N/A 

N/A 

N/A 

N/A 

+ 

N/A 

N/A 

N/A 

N/A 

N/A 

N/A 

N/A 

N/A 

+ 

+ 

N/A 

N/A 

+ 

Finally, turning to specific verb choice in the L1 and L2 data reveals some trends which 

could be particularly helpful for teachers and textbook/materials developers. First, there were a 

few verbs that had similar meanings that appeared in the learner data, when another (perhaps 

more formal or academic) form appeared in the L1 data. One example of this is the difference 

between the verb sal (live) and geojuha (reside). While functionally similar, sal was found to be 

a distinctive collexeme for the progressive -ko iss in all varieties. On the other hand, geojuha was 

only found to be a distinctive collexeme in the L1 data. This difference exemplifies how learners 

may tend to rely on more common/less academic language, which is something pedagogues 

should be aware of. Another example of this was the pair of verbs jeunggaha (increase) and 

manhaji (increase). While both verbs have the same semantic meaning, manhaji was only found 

to be a distinctive collexeme in the L1 English learner data. In the L1 Korean data, only 

jeunggaha and neuleona (increase) were found. One possible explanation is that as manhaji is 

73 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
used more commonly in speech and jeunggaha appears frequently in writing (such as articles or 

news reports) that learners have more frequent exposure to manhaji. As such, learners can 

benefit from being made aware of how semantically similar verbs in Korean are used differently 

depending on the modality. Distinctive collexeme analysis allows the analyst to quantitatively 

identify such variation between L1 and L2 language as discussed in this section. 

In this section, I have covered the results from the distinctive collexeme analysis for three 

varieties: L1 Korean, L1 English L2 Korean, and L1 Japanese L2 Korean. I have discussed how 

the usage of the (non)progressive appears to vary across varieties, highlighting both key 

similarities and differences as it pertains to potential benefits for language teachers and materials 

developers. Additionally, despite this study being an exploratory corpus study in nature, I have 

offered some discussion of potential interlanguage transfer effects which may be impacting the 

usage of the progressive in the learner varieties, and highlighted how even learners who speak a 

language which is typologically dissimilar from Korean exhibit the ability to acquire and using 

distinctly Korean constructions (e.g., L1 English speakers usage of the progressive with the verb 

know). In what follows, I investigate the usage of the progressive in each variety using regression 

analysis to see if any identified predictors, or their interactions, appear to have an impact on the 

choice to use a (non)progressive.  

74 

 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.4 

Normalized coll.strengths in L1 Korean data. 

Figure 4.5 

Normalized coll.strengths in L1 English L2 Korean data. 

75 

 
 
 
 
 
 
 
Figure 4.6 

Normalized coll.strengths in L1 Japanese L2 Korean data. 

4.3. Research question 2 

4.3.1. Textbook analysis of -ko iss 

The number of verbs appearing in the progressive -ko iss construction increased as the textbook 

level increased, as shown in Table 4.2, likely due to the increasing length and complexity of the 

texts featured in each series as the level progressed. In terms of the most frequent verbs featured 

in each series, there was some overlap across textbook series. For example, included in the top 

five most frequent verbs were sal (live), the most frequent verb used with the progressive -ko iss 

in both textbook series. Notably, the stative verb al (know) was also included in the top verbs of 

both textbook series (ranked #4 in the New Sogang Korean series and #2 in the KLEAR 

Integrated Korean series). In addition to al (know), the stative verb gaji (have/hold) was 

included in the Integrated Korean series as the fifth most frequently featured verb in the 

progressive overall.  

76 

 
 
 
Table 4.2.  

Raw frequency of the -ko iss construction in L2 Korean textbooks. 

Level 1 

Level 2 

Level 3 

Level 4 

Total 

Textbook 1 

Textbook 2 

0 

18 

23 

25 

62 

99 

100 

147 

185 

289 

Table 4.3. 

Raw frequency and proportions of verbs appearing in the progressive -ko iss construction in 

New Sogang Korean and KLEAR Integrated Korean by level. 

New Sogang Korean 

KLEAR Integrated Korean 

Verb 

Frequency (%)  Verb 

Frequency (%) 

Level 1 

– 

– 

– 

– 

– 

– 

– 

– 

– 

– 

– 

– 

– 

– 

ha- ‘do’ 

ta- ‘ride’ 

baeu- ‘learn’ 

dani- ‘attend’ 

deud- ‘listen’ 

Others 

Total 

Level 2 

baeu- ‘learn’  

2 (8.70) 

sal- ‘live’ 

junbiha- ‘prepare 

2 (8.70) 

chaj- ‘find’ 

sal- ‘live 

2 (8.70) 

jinae- 
‘spend/pass’ 

4 (22.22) 

3 (16.67) 

1 (5.56) 

1 (5.56) 

1 (5.56) 

8 (44.44) 

18 (100) 

3 (12) 

2 (8) 

2 (8) 

yaegiha- ‘talk’ 

2 (8.70) 

gaj- ‘have/hold’ 

2 (8) 

77 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 4.3 (cont’d). 

bo- ‘see’ 

Others 

Total 

1 (4.35) 

al- ‘know’ 

14 (60.87) 

Others 

23 (100) 

Total 

1 (4) 

15 (60) 

25 (100) 

Level 3 

chaj- ‘find’ 

7 (11.29) 

sal- ‘live’ 

11 (11.11) 

baeu- ‘learn’ 

gidari- ‘wait 

sal- ‘live’ 

ul- ‘cry’ 

Others 

Total 

5 (8.06) 

al- ‘know’ 

5 (8.06) 

jinae- 
‘spend/pass’ 

5 (8.06) 

ga- ‘go’ 

3 (4.84) 

ggeul-‘pull/draw in’ 

5 (5.05) 

4 (4.04) 

3 (3.03) 

3 (3.03) 

37 (59.68) 

Others 

73 (73.74) 

62 (100) 

Total 

99 (100) 

7 (4.76) 

Level 4 

sal- ‘live’ 

7 (7.00) 

sal- ‘live’ 

al- ‘know’ 

4 (4.00) 

gaji- ‘have/hold’ 

7 (4.76) 

iyagiha- ‘talk’ 

4 (4.00) 

ga- ‘go’ 

bo- ‘see’ 

3 (3.00) 

gaji- ‘have/hold’ 

3 (3.00) 

jui- ‘take 
control’* 

jaesiha- 
‘suggest’ 

6 (4.08) 

5 (3.40) 

5 (3.40) 

Others 

Total 

79 (79.00) 

Others 

117 (79.50) 

100 (100) 

Total 

147 (100) 

*쥐다 can also mean have/hold/squeeze, however, its use in Textbook 2-Level 4 related to taking control of 
economic power or finances (e.g., 한국의가정에서경제권을쥐고있는사람이누구일까요? In Korean homes, 
who is the one who is has/takes control over finances?) 

78 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 4.4.  

Five most frequent verbs used with the progressive -ko iss construction across all volumes of 

New Sogang Korean and KLEAR Integrated Korean 

New Sogang Korean 

KLEAR Integrated Korean 

Verb 

Frequency # (%)  Verb 

Frequency # (%) 

sal- ‘live’ 

14 (7.67) 

sal- ‘live’ 

chaj- ‘find’ 

9 (4.86) 

al- ‘know’ 

baeu- ‘learn’ 

8 (4.32) 

ga- ‘go’ 

gidari- ‘wait’ 

8 (4.32) 

gaji- 
‘have/hold’ 

22 (7.61) 

11 (3.80) 

9 (3.11) 

9 (3.11) 

al- ‘know’ 

6 (3.24) 

ha- ‘to do’ 

8 (2.77) 

1 

2 

3 

4 

5 

4.3.2. Teaching of the progressive -ko iss in the textbooks 

In addition to determining the frequencies of verbs appearing in the progressive across textbook 

levels, I also qualitatively explored how the progressive is taught and introduced in each series. 

In both series, the progressive is introduced early on. In the New Sogang textbook series the 

progressive -ko iss is introduced in volume 2A, the third volume. In Integrated Korean, the 

construction is introduced in Beginner 2 (the second volume). In both series, the construction is 

only explicitly taught with verbs that denote physical actions, such as bo (watch), deud (listen), 

cheongsoha (clean), drink (masi), and make (mandeul), among others. Following the 

introduction of the progressive, both texts incorporate short dialogues that demonstrate the usage 

of the progressive. Below are short excerpts from both New Sogang Korean and Integrated 

Korean, respectively: 

79 

 
 
 
 
 
 
New Sogang Korean:  

Excerpt from 2 과말하기대화1 (Lesson 2 Speaking Dialogue 1): 전화를받을수없는

이유설명하기 – Explaining the reason you cannot take a phone call. 

제니:앤디씨,지금통화할수있으세요?

Jenny: Andy, can you call now? 

앤디:미안해요.제가지금친구하고얘기하고있어요.

Andy: Sorry. I’m talking with my friend now. 

KLEAR Integrated Korean:  

Excerpt from Integrated Korean: Beginning Two:  Conversation 1 (차한잔마실래요? – 

Shall we have a cup of tea?) 

유진:어,민지씨아니세요?뭐하세요?

Yujin: Hey, aren’t you Minji? What are you doing? 

민지:차마시고있어요.우진씨도차한잔하실래요?

Minji: I’m drinking tea. Yujin, would you like a cup of tea, too? 

유진:네,저도마시고싶었는데잘됐네요.

Yujin: Yes, I also wanted to drink one so this worked out well. 

In both textbooks, the -ko iss construction is introduced in English as a construction used to 

denote a continuous action or an action in progress. In the New Sogang Korean series the 

construction is defined as follows: 

Meaning:‘-고있다’ is used to express actions in progress or repeated actions. It has the 

same meaning as “to be doing (something”.  

80 

 
Form:‘고있다' is always attached directly to the verb stem.’  

Likewise, in Integrated Korean, the construction is introduced as the following: 

~고있다 expresses the continuation or progression of an action. Only verbs (not 

adjectives) can occur in this construction.  

Of note in both textbooks is that the discussion of the usage of -ko iss outside of the ‘action in 

progress’ senses are not explicitly discussed. Usage of the progressive with stative and mental 

verbs comes in later volumes and is incorporated in readings and dialogues as it is used in daily 

conversation or written texts, though at low frequencies. For example, New Sogang Korean 

includes al (know) in the progressive when a character in a dialogue is talking to their friend 

whose dream to become a news anchor came true (from New Sogang Korean, volume 4B): 

Korean:나는네가유명한앵커가될거라는것을10 년전에미리알고있었어.

English: I knew even ten years ago that you would become a famous news anchor.  

As can be seen from the example from New Sogang Korean, al (know) and its use with the 

progressive -ko iss is represented. As the form is common in spoken Korean, the textbook 

incorporates it in the informal spoken form.  

Examples of stative verbs in their written forms are also represented, for example, the 

verb gaji (have/hold) appears in the textbooks as a mental and stative verb to describe having or 

holding a meaning. Of note here, as well, is that we see the -ko iss with a stative progressive 

being used with an inanimate subject. Example below is taken from Integrated Korean: High 

Intermediate II: 

81 

 
 
 
Korean: ‘바쁜사람들도,굳센사람들도,바람과같던사람들도,집에돌아오면

아버지가된다.’

‘연예인이름보다꽃이름을더많이아는아이로키우고싶습니다.’

이러한광고카피에는소비자들의많은호응과칭찬이이어졌다.이렇듯현대의

아파트는우리가사는주거공간이상의의미를가지고있다.

English: "The busy, the strong, like the wind, when they return home, they become 

fathers." 

"I want to raise him to know more about flower names than celebrities." 

This advertising copy was followed by a lot of response and praise from consumers. As 

such, modern apartments have more meaning than just the residential space we live in. 

In the example above, gaji is used to express how apartments have or hold (sentimental) meaning 

in that they are important places for people and families to grow up in, with gaji (have/hold) 

being marked with the progressive -ko iss. This article in the textbook was describing apartments 

in Korea, and how Korean people generally prefer to live in apartments, and how apartments are 

advertised. Notably, this is a prime example of the textbooks using the -ko iss construction in a 

way exhibited in Korean, with an inanimate subject and a stative progressive. 

Given the above, this analysis shows how -ko iss is used with a limited variety of verbs 

overall. It also shows that, despite each series only teaching the prototypical usage of the 

progressive explicitly, in later volumes of each series, -ko iss is incorporated with certain mental 

verbs such as al (know). That being said, the frequency analysis reveals that, overall, the 

progressive’s usage in the textbooks is relatively low, with no more than a few hundred examples 

over the course of two textbook series and sixteen volumes. The progressive is used in a variety 

82 

 
 
of genres in the textbooks, though, including conversational and casual dialogue and articles, 

offering learners the opportunity to notice the form -ko iss with semantic meanings and usage 

cases beyond the prototypical ‘action in progress’ usage.  

As far as grammar description goes, the textbooks appear to lack explicit instruction on 

the form-function mappings of the -ko iss construction as it is used beyond simply describing 

actions in progress and with stative and mental verbs. Including (i) more examples of the 

progressive in such usage cases and (ii) offering a description of how -ko iss can co-occur with 

stative and mental verbs, and be used with inanimate subjects, could help learners acquire and 

use the form in a way that is more in line with L1 speakers of Korean. As the prototypical -ko iss 

description appears early in the textbooks, textbook developers can incorporate descriptions of -

ko iss beyond ‘action in progress’ starting from the intermediate textbook series and include 

explicit descriptions of how it is used in both spoken and written language. 

4.3.3. Comparing L1 and L2 corpora with textbooks 

To compare verbs appearing in the corpora (L1 and L2) and textbooks, I normalized the 

frequencies of key verbs in each to better compare general trends in the usage of each verb type 

based on variety. Figures 4.7, 4.8, and 4.9 visually represent the normalized frequencies of a few 

key verbs. First, turning attention to the most common verb associated with the progressive in 

the textbook data, sal (live), it is clear that learners use the verb in the progressive in their writing 

much more than the verb appeared in the L1 corpus with the progressive. L1 English speakers 

used it about 5.5 times more frequently, and L1 Japanese speakers used it about 7.4 times as 

frequently. As Japanese allows for the verb of the same meaning (sumu, to live) to be used 

commonly with the progressive, it is not surprising that the Japanese data presents the 

83 

 
progressive construction being used commonly with sal (to live) in Korean. Also of note is that 

the verb is similarly frequent in both textbook series. 

Moving on to the stative verbs that appeared in the top five most frequently used verbs in 

the textbooks, I explored gaji (have/hold) and al (know). Gaji can be used to express holding or 

having physical objects, but it can also be used to express having or holding opinions, thoughts, 

or feelings, and this more abstract usage might be difficult for learners to understand and acquire. 

However, the normalized frequencies suggest that, indeed, both varieties of learner language 

studied here exhibit a usage of this verb in the progress construction, with their normalized 

frequencies coming in at higher than the normalized frequencies of the verb in the L1 corpus. 

English learners of Korean used gaji with the progressive construction 4.5 times more, and the 

Japanese learners of Korean used gaji with the progressive construction 5.4 times more 

frequently than it was used in the L1 corpus. Again, the progressive usage with a stative verb is 

higher in the L1 Japanese data, perhaps as expected due to the typological similarities between 

Japanese and Korean. 

Of important note is al (know), which is a mental and stative verb which commonly takes 

the progressive construction in Korean. Al was ranked fifth in New Sogang Korean series, and 

second in KLEAR Integrated Korean series in terms of its rate of co-occurrence with the 

progressive. Al is notable, as it is a prime example of how textbook frequencies differ from real-

world usage frequencies. Figure 4.9 shows the normalized frequencies of al, it is obvious that its 

usage in the L1 corpus surpasses its usage in both textbook series, with al appearing 2.6 times 

less frequently in New Sogang Korean, and 2.1 times less frequently in KLEAR Integrated 

Korean, suggesting an underuse of the construction in the textbooks when compared with real-

world corpus data. L1 English learners of L2 Korean exhibited a perhaps surprisingly high rate 

84 

 
 
 
of usage of al with the progressive, though their usage rate was still about 1.5 times less than the 

L1 data. However, given that English rarely, if ever, allows for the verb know to appear in the 

English progressive be… ing construction, this is a positive finding for English speaking learners 

of Korean. For L1 Japanese learners, the trends are different, as their usage of the progressive 

with al is about 1.4 times higher than the L1 data, 2.14 times higher than their L1 English 

counterparts, and 3.6 times higher than New Sogang Korean and 2.6 times higher than KLEAR 

Integrated Korean. As Japanese allows for stative progressives, and in particular, allows for the 

Japanese verb know (shiru) to co-occur frequently with the Japanese progressive construction (-

te iru), there is evidence that even in the case of relatively sparse representation in the textbook 

data, a learner’s L1 will play a role in their target-like usage of a construction. In this case in 

particular, that is borne out with the comparisons of the L1 English group underuse the verb with 

the progressive, and the L1 Japanese group using the construction more frequently. Overall, this 

shows a clear need for textbooks to be designed with their target audience in mind, for example, 

a textbook geared towards L1 English speakers may need more examples of al with -ko iss to 

help learners notice the form. 

85 

 
 
 
 
 
 
 
 
 
Figure 4.7. 

Relative frequency of lemma sal (live) with -ko iss across corpora. 

86 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.8. 

Relative frequency of lemma gaji (live) with -ko iss across corpora. 

87 

 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 4.9. 

Relative frequency of lemma al (know) with -ko iss across corpora. 

4.4. Research question 3 

To assess what explanatory variables may influence the choice of a progressive -ko iss or a non-

progressive in the L1 and L2 data, I chose to run a binary logistic regression in JASP (2024) 

version 0.18.3. Logistic regression is used when the dependent variable consists of two possible 

outcomes. In this case, whether the construction choice is (a) the progressive -ko iss or (b) the 

non-progressive. In logistic regression, the predictors (explanatory variables) are categorical or 

scale (Brezina, 2018). Logistic regression can be of particular use in corpus linguistics as the 

method itself does not assume a linear relationship between the dependent and independent 

variables, nor do they need to be normally distributed or of equal variance within each group. 

Likewise, the residuals do not need to be normally distributed. However, the dependent variable 

must be dichotomous, and the categories for the dependent variable must be exhaustive in that 

88 

 
 
 
every case submitted to the binary logistic regression must be a member of only one group (in 

other words, either progressive or non-progressive). Multicollinearity must be checked to ensure 

that no predictor variables are highly correlated with each other, and this is assessed by checking 

the VIF, or Variance Inflation Factor3, multicollinearity diagnostics in JASP. As this is an 

exploratory analysis, entry method is used.  

Approximately 500 random samples from each variety (L1 Korean, L1 English L2 

Korean, L1 Japanese L2 Korean) were extracted from the overall dataset for manual annotation 

of predictors, totaling 1523 extractions. Random sampling was done using the randomize range 

feature in Google Sheets. In addition to variety and construction, the data were manually 

annotated for aktionsart (four levels: activity/process, accomplishment, achievement, stative), 

semantic domain (seven levels: activity, communication, mental, causative, occurrence, 

existence, aspectual), and animacy of the subject (three levels: animate, human, inanimate). In 

the present study, I did not include speaker/author as a fixed effect because in the L1 corpus that 

information is not known, though I acknowledge that the inclusion of such fixed effects can help 

in building the best model. In total, I completed approximately 4500 manual annotations.  

According to Brezina (2018), block entry is “usually preferable” in corpus studies 

employing logistic regression as the predictor variables to include have been decided based on 

literature or theory (p. 123). As the predictor variables for the present study were chosen based 

on existing literature, I employed the block entry method when running the regression in JASP.  

3 Variance Inflation Factor (VIF) is a statistic used to check for multicollinearity between predictor variables. 
Generally, VIFs larger than 10 are considered as a warning sign of multicollinearity issues. See 
https://online.stat.psu.edu/stat462/node/180/ for a discussion on VIFs. 

89 

 
  
 
 
4.4.1. Inter-rater reliability 

Approximately 10% of the annotations (about 450) were separately annotated by an L1 speaker 

of Korean for inter-rater reliability. The rater hired for this study, at the time of participation, 

held an advanced degree in linguistics and language education from a Korean university and was 

teaching Korean language courses in the North American University context. The rater was 

reimbursed at a rate of $20 USD per hour.  

Reliability statistics were calculated using JASP, with the reliability function installed. 

Cohen’s kappa was calculated for each explanatory variable separately, and the output was 

interpreted considering guidelines for interpreting Cohen’s kappa put forward by Landis and 

Koch (1977). For aktionsart, Cohen’s kappa was .68, showing substantial agreement. For 

animacy, Cohen’s kappa was .79, showing substantial agreement. For semantic domain, Cohen’s 

kappa was .69, showing substantial agreement. Variety and construction did not require inter-

rater agreement statistics. Given the adequate inter-rater reliability statistics between both raters, 

the data was further analyzed. Due to time constraints, the hired annotator and I were unable to 

discuss and re-annotate for areas of disagreement. For the present analysis, my annotations are 

used.  

4.4.2. Results 

Overview summary: As this dissertation is exploratory with the goal of identifying what factors 

may contribute to the choice to use a progressive in L1 and L2 Korean, I ran three models, with 

the goal of improving the explanatory power of the model each time. I will briefly summarize 

each model: For the first model, model 1, each predictor variable was added to the first model 

without interactions (block entry was used for all models). To investigate whether the influence 

of the L1 of the writer combined with other predictor variables (aktionsart, semantic domain, 

90 

 
 
animacy) influence the choice to use a progressive or non-progressive, I introduced interactions 

between variety and aktionsart, variety and semantic domain, and variety and animacy in model 

2. While this model did show some statistically significant interactions, I found it had issues with 

Standard Errors larger than the estimates for some interactions between semantic domain and 

variety. Following Brezina’s guidelines which warn that Standard Errors larger than the 

estimates suggest something is wrong with the model, I then explored the predictor semantic 

domain to identify the cause using contingency tables. I identified that extremely low rates of the 

aspectual, causative, and communication levels of semantic domain appeared to be causing this 

issue. I then ran a third model sans those levels of semantic domain, which remedied the issue 

with Standard Error. This model, model 3, will be used to discuss potential interactions that may 

influence the choice of the progressive in L1 and L2 Korean writing. 

4.4.3. Logistic regression – first model 

Multicollinearity was assessed for each of the explanatory variables by calculating the Variance 

Inflation Factor (VIF) for each explanatory variable in the logistic regression. Generally, a VIF 

of greater than 10 indicates multicollinearity. No explanatory variable had a VIF of greater than 

10 (semantic domain: 3.36; animacy: 2.17; aktionsart: 2.08; variety: 1.58). I also used a 

Confusion Matrix as a performance diagnostic in JASP to assess the overall accuracy of the 

model’s predictions. The Confusion Matrix yielded an Overall Correct Prediction Rate of 

69.96%, indicating that the model is correctly predicting about seventy percent of the time.  

Model 1 was statistically significant (p = .001) with Nagelkerke R2 of 0.241 (Nagelkerke 

effect size computed between 0 and 1), so about 24.1% of the variance in the dependent variable 

can be explained by the model. Summaries of model 1 are included in Table 4.1 and Table 4.2. 

91 

 
Several levels of semantic domain were significant predictors in the model, (aspectual: 

OR = .18, p < .001; communication: OR = .54, p < .008; existence: OR = .31,  p < .001; mental: 

OR = .18 p < .001). Verbs in these categories appear more likely to be used in the non-

progressive, or show a dispreference for the progressive -ko iss. The factor of animacy was also 

found to be statistically significant on the level of human (OR = .22, p < .001). 

For aktionsart verb categorizations, statistically significant results were found for 

achievement verbs (OR = 0.518; p = .007), activity verbs (OR = 1.92; p = .003), and stative 

verbs (OR = 1.829; p = .009). These results show that when a verb falls into the activity or 

stative aktionsart category, it is likely to trigger use of -ko iss (progressive) in Korean. Likewise, 

when a verb falls into the achievement category, it is less likely to trigger a progressive.  

Table 4.1 

Model 1 summary. 

Model Summary 

Model  Deviance 
H₀ 
H₁ 

2103.720   2105.720   2111.045   1517  
1801.074   1829.074   1903.626   1504   302.646   < .001    

AIC 

BIC 

df 

Χ² 

p 

Nagelkerke R² 

0.241      

Table 4.2. 

Results for model 1. 

92 

 
 
  
 
   
      
     
 
 
 
 
 
 
 
Coefficients  

Wald Test 

(Intercept) 
aktionsart (achievement) 
aktionsart (activity) 
aktionsart (stative) 
animacy (human) 
animacy (inanimate) 
semantic_domain (aspectual) 
semantic_domain (causative) 
semantic_domain 
(communication) 
semantic_domain (existence) 
semantic_domain (mental) 
semantic_domain (occurrence) 
variety (L2_ENG) 
variety (L2_JPN) 

Estimate 
  0.841 
  -0.657   
  0.656 
  0.604 
  -1.516   
  -0.104   
  -1.700   
  0.666 
  -0.615   
  -1.176   
  -1.495   
  -0.022   
  0.517 
  0.565 

Standard 
Error 
0.273 
0.242 
0.218 
0.232 
0.232 
0.236 
0.398 
0.563 

0.230 

0.330 
0.181 
0.207 
0.169 
0.164 

Odds 
Ratio 
  2.319 
  0.518 
  1.928 
  1.829 
  0.220 
  0.901 
  0.183 
  1.946 
  0.541 
  0.309 
  0.224 
  0.978 
  1.677 
  1.759 

z 
  3.080  
 -2.716  
  3.013  
  2.605  
 -6.539  
 -0.440  
 -4.270  
  1.183  
 -2.669  
 -3.559  
 -8.266  
 -0.107  
  3.063  
  3.442  

Wald 
Statistic 
9.485 
7.378 
9.076 
6.787 
42.759 
0.193 
18.237 
1.400 

7.125 

12.670 
68.325 
0.011 
9.382 
11.850 

Note.  progressive level 'yes' coded as class 1. 

4.4.4. Logistic regression – second model 

95% Confidence interval  
(odds ratio scale) 
Upper 
Lower 
bound 
bound 
3.961 
1.358 
0.833 
0.323 
2.954 
1.258 
2.881 
1.161 
0.346 
0.139 
1.431 
0.568 
0.399 
0.084 
5.866 
0.646 

0.344 

0.161 
0.157 
0.653 
1.205 
1.275 

0.849 

0.590 
0.320 
1.466 
2.334 
2.427 

df  p 
 1   0.002  
 1   0.007  
 1   0.003  
 1   0.009  
 1  < .001  
 1   0.660  
 1  < .001  
 1   0.237  
 1   0.008  
 1  < .001  
 1  < .001  
 1   0.915  
 1   0.002  
 1  < .001  

To explore potential interaction effects, a second logistic regression, model 2, was run. 

Theoretically, it is assumed that variation could exist between L1 Korean and L2 Korean-English 

and Korean-Japanese varieties due to typological differences. As such, in the second logistic 

regression, I used JASP to explore interactions between the predictors and variety (i.e., L1 and 

Learner Language). The summary and results for model 2 are listed in Table 4.3 and Table 4.4.  

Table 4.3 

Model 2 summary. 

Model  Deviance 
H₀ 
H₁ 

2103.720   2105.720   2111.045   1517  
1692.669   1764.669   1956.375   1482   411.051   < .001 

AIC 

BIC 

df 

Χ² 

p 

Nagelkerke R² 

0.316 

93 

 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
   
  
   
 
 
  
 
 
 
 
 
 
Table 4.4 

Results for model 2. 

Coefficients  

Estimate 

(Intercept) 
aktionsart 
(achievement) 
aktionsart 
(activity) 
aktionsart (stative) 
animacy (human) 
animacy 
(inanimate) 
semantic_domain 
(aspectual) 
semantic_domain 
(causative) 
semantic_domain 
(communication) 
semantic_domain 
(existence) 
semantic_domain 
(mental) 
semantic_domain 
(occurrence) 
variety (L2_ENG) 
variety (L2_JPN) 
semantic_domain 
(aspectual) * 
variety (L2_ENG) 
semantic_domain 
(causative) * 
variety (L2_ENG) 
semantic_domain 
(communication) * 
variety (L2_ENG) 
semantic_domain 
(existence) * 
variety (L2_ENG) 

Wald Test 

95% Confidence interval 

SE 

0.353 

0.323 

0.312 

0.315 
0.283 

0.277 

0.449 

0.681 

0.312 

0.440 

0.344 

0.314 

1.067 
0.750 

Odds Ratio 

3.158 

0.614 

1.137 

1.069 
0.341 

0.917 

z 
  3.255 
 -1.510   

  0.411 
  0.211 
 -3.791   
 -0.314   

Wald 
Statistic 
10.593 

2.280 

df 
p 
 1   0.001   
 1   0.131   

0.169 

0.045 
14.375 

0.098 

 1   0.681   
 1   0.833   
 1  < .001   
 1   0.754   

Lower 
bound 
0.457 

-1.122 

-0.484 

-0.550 
-1.630 

-0.630 

0.153 

 -4.182   

17.491 

 1  < .001   

-2.757 

2.351 

  1.255 

1.575 

 1   0.210   

-0.480 

0.337 

 -3.493   

12.199 

 1  < .001   

-1.700 

0.327 

 -2.540   

6.452 

 1   0.011   

-1.982 

0.473 

 -2.175   

4.730 

 1   0.030   

-1.423 

0.421 

2.179 
0.750 

 -2.753   
  0.730 
 -0.384   

7.579 

0.533 
0.147 

 1   0.006   
 1   0.465   
 1   0.701   

-1.482 

-1.313 
-1.759 

Upper 
bound 
1.842 

0.146 

0.740 

0.683 
-0.519 

0.456 

-0.998 

2.190 

-0.478 

-0.255 

-0.074 

-0.249 

2.871 
1.182 

1.150   
-0.488   

0.128   
0.066   
-1.075   
-0.087   

-1.877   

0.855   

-1.089   

-1.119   

-0.748   

-0.866   
0.779   
-0.288   

14.416   

1455.398 

  1.824×10+6    0.010 

  9.812×10-5  1   0.992    -2838.111 

  2866.944 

-2.958   

1.408 

0.052 

 -2.100   

4.411 

 1   0.036   

-5.717 

-0.198 

16.083   

467.663 

  9.653×10+6    0.034 

0.001 

 1   0.973   

-900.520 

  932.686 

-1.767   

1.067 

0.171 

 -1.656   

2.741 

 1   0.098   

-3.859 

0.325 

94 

 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 4.4 (cont’d). 

semantic_domain 
(mental) * variety 
(L2_ENG) 
semantic_domain 
(occurrence) * 
variety (L2_ENG) 
semantic_domain 
(aspectual) * 
variety (L2_JPN) 
semantic_domain 
(causative) * 
variety (L2_JPN) 
semantic_domain 
(communication) * 
variety (L2_JPN) 
semantic_domain 
(existence) * 
variety (L2_JPN) 
semantic_domain 
(mental) * variety 
(L2_JPN) 
semantic_domain 
(occurrence) * 
variety (L2_JPN) 
aktionsart 
(achievement) * 
variety (L2_ENG) 
aktionsart 
(activity) * variety 
(L2_ENG) 
aktionsart (stative) 
* variety 
(L2_ENG) 
aktionsart 
(achievement) * 
variety (L2_JPN) 
aktionsart 
(activity) * variety 
(L2_JPN) 
aktionsart (stative) 
* variety (L2_JPN) 
animacy (human) * 
variety (L2_ENG) 
animacy 
(inanimate) * 
variety (L2_ENG) 
animacy (human) * 
variety (L2_JPN) 
animacy 
(inanimate) * 
variety (L2_JPN) 

-1.061   

0.456 

0.346 

 -2.326   

5.412 

 1   0.020   

-1.955 

-0.167 

1.411   

0.588 

4.100 

  2.401 

5.766 

 1   0.016   

0.259 

2.563 

15.761   

1026.010 

  6.998×10+6    0.015 

  2.360×10-4  1   0.988    -1995.182 

  2026.704 

12.865   

1026.476 

  386726.757    0.013 

  1.571×10-4  1   0.990    -1998.991 

  2024.721 

0.450   

0.541 

1.568 

  0.833 

0.693 

 1   0.405   

-0.609 

1.510 

1.277   

1.013 

3.585 

  1.260 

1.589 

 1   0.208   

-0.709 

3.262 

-1.254   

0.506 

0.285 

 -2.479   

6.143 

 1   0.013   

-2.245 

-0.262 

2.074   

0.582 

7.956 

  3.566 

12.719 

 1  < .001   

0.934 

3.214 

-1.372   

0.750 

0.254 

 -1.830   

3.350 

 1   0.067   

-2.842 

0.097 

0.545   

0.646 

1.725 

  0.845 

0.714 

 1   0.398   

-0.720 

1.811 

1.032   

0.671 

2.805 

  1.538 

2.365 

 1   0.124   

-0.283 

2.346 

0.044   

0.627 

1.045 

  0.070 

0.005 

 1   0.944   

-1.184 

1.272 

0.965   

0.528 

2.624 

  1.827 

3.337 

 1   0.068   

-0.070 

2.000 

0.907   

-1.178   

0.588 

0.962 

2.476 

  1.541 

2.375 

 1   0.123   

-0.246 

0.308 

 -1.224   

1.499 

 1   0.221   

-3.063 

2.060 

0.708 

-0.008   

0.974 

0.992 

 -0.008    6.289×10-5  1   0.994   

-1.916 

1.901 

-0.203   

0.639 

0.816 

 -0.318   

0.101 

 1   0.750   

-1.455 

1.049 

-0.130   

0.717 

0.878 

 -0.181   

0.033 

 1   0.856   

-1.534 

1.275 

Note.  progressive level 'yes' coded as class 1. 

95 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
However, a look at the model summary reveals some critical issues with model 2. While 

the model does show some statistically significant interactions, the model ultimately suffered 

from high Standard Errors for several interactions and several confidence intervals spanning 1, 

which serve as a warning sign that the model 2 has some issues. Looking at the output, issues 

seem to occur when variety interacts with semantic domain on the level of aspectual, causative, 

and communication.  

To address this, I took an exploratory approach using contingency tables in JASP to see 

the distribution of the annotations for semantic domain, taking care to explore the distribution of 

annotations in each variety (L1 and L2) separately. This analysis revealed that across L1 and L2 

there were lower numbers of exemplars annotated as aspectual, causative, and communication. 

For example, in the L1 data, only 14 verbs co-occurring with -ko iss were annotated as causative, 

and likewise only ten verbs co-occurring with -ko iss were annotated as aspectual. L1 Japanese 

L2 Korean data followed a similar trend, with only two verbs co-occurring with -ko iss annotated 

with aspectual and causative semantic domains each; only nine verbs received an existence 

annotation for semantic domain. L1 English L2 Korean data, likewise, had one verb in -ko iss for 

aspectual, two for causative, and one for existence. To attempt to remedy this issue, I removed 

the levels of aspectual, causative, and communicative from the data and ran a final model to 

attempt to find potential interactions. 

96 

 
 
 
 
 
 
 
 
Table 4.5 

Contingency table for levels of semantic domain in L1 Korean. 

Contingency Tables  

progressive 

Semantic domain  no 
42  
activity 
27  
aspectual 
3  
causative 
50  
communication 
20  
existence 
51  
mental 
56  
occurrence 
Total 

yes  Total 
77   119  
10   37  
14   17  
32   82  
18   38  
42   93  
57   113  
  249   250   499  

Table 4.6 

Contingency table for levels of semantic domain in L1 English L2 Korean. 

Contingency Tables  

progressive 

Semantic domain  no 
activity 
aspectual 
causative 
communication 
existence 
mental 
occurrence 
Total 

yes  Total 
  117   129   246  
1  
1  
2  
4  
9  
9  
2   10  
44   155  
78   89  
  249   265   514  

0  
2  
0  
8  
  111  
11  

97 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table 4.7 

Contingency table for levels of semantic domain in L1 Japanese L2 Korean. 

Contingency Tables  

progressive 

Semantic domain  no 
activity 
aspectual 
causative 
communication 
existence 
mental 
occurrence 
Total 

yes  Total 
71   121   192  
2  
2  
0  
2  
0  
2  
13   26  
13  
9   11  
2  
39   191  
  152  
74   85  
11  
  249   260   509  

4.4.5. Logistic regression – third model 

Because the focus of this section was to attempt to identify potential interactions between variety 

and semantic domain, variety and aktionsart, and variety and animacy, these predictors were put 

into model 3 using entry method; interactions were added to the model manually in JASP. 

Overall, model 3 is an improvement over model 2 and allows for cautious optimism in its 

interpretation. The model 3 is statistically significant with a low p-value (p < .001) and decent 

effect size (Nagelkerke R2 = .32). Similar to model 1, model 3 shows that semantic domains on 

their own seem to predict the use of a non-progressive: existence (OR = .35, p = .02), mental 

(OR = .45, p = .02), occurrence (OR = .45, p = .01). Interaction effects were found between 

variety and semantic domain. According to this model, L1 English speakers are less likely to use 

a progressive when the verb’s semantic domain is mental (OR = .37, p = .03). L1 Japanese 

speakers follow a similar pattern in terms of a verb’s semantic domain being categorized as 

mental (OR = .28, p .01). Both L1 English and L1 Japanese learners of L2 Korean appear to be 

more likely to use a progressive when the semantic domain of the verb is occurrence based on 

98 

 
 
 
 
 
 
 
 
 
 
 
the interactions between Variety (English)*Semantic domain (occurrence) and Variety 

(Japanese)*Semantic domain (occurrence) (L1 English: OR = 3.5, p = .04; L1 Japanese: OR = 

8.00, p < .001). L1 English learners of L2 Korean are more less likely to use a progressive when 

the subject is animate (OR = .09, p = .05).  

While these results may scratch the surface on what aspects of a verb (phrase) may 

influence the choice of a progressive or non-progress across L1 and L2 Korean, even model 3 

has some shortcomings which cannot go unstated. First, while model 3 was able to address 

shortcomings that plagued model 2, such as the high Standard Errors, many Confidence Intervals 

in model 3 were found to include 1, which suggests results may not be statistically significant. 

This highlights the difficulty of modeling corpus data, particularly when incorporating 

interactions. Going forward, the way to address this would be to (i) include more data from each 

variety and (ii) reconsider some categories for annotation. This is discussed in more detail in 

section 5.2.1 (Future directions). 

Table 4.8 

Model 3 summary. 

Model  Deviance 
H₀ 
H₁ 

  1853.543    1855.543    1860.742   1337  
  1492.387    1546.387    1686.758   1311   361.156    < .001 

AIC 

BIC 

df 

Χ² 

p 

Nagelkerke R² 

0.316 

99 

 
 
 
 
 
 
     
   
 
 
 
 
 
 
 
Table 4.6 

Results for model 3. 

Coefficients  

(Intercept) 
variety (L2_ENG) 
variety (L2_JPN) 
aktionsart (achievement) 
aktionsart (activity) 
aktionsart (stative) 
semantic_domain (existence) 
semantic_domain (mental) 
semantic_domain (occurrence) 
animacy (human) 
animacy (inanimate) 
variety (L2_ENG) * aktionsart 
(achievement) 
variety (L2_JPN) * aktionsart 
(achievement) 
variety (L2_ENG) * aktionsart (activity) 
variety (L2_JPN) * aktionsart (activity) 
variety (L2_ENG) * aktionsart (stative) 
variety (L2_JPN) * aktionsart (stative) 
variety (L2_ENG) * semantic_domain 
(existence) 
variety (L2_JPN) * semantic_domain 
(existence) 
variety (L2_ENG) * semantic_domain 
(mental) 
variety (L2_JPN) * semantic_domain 
(mental) 
variety (L2_ENG) * semantic_domain 
(occurrence) 
variety (L2_JPN) * semantic_domain 
(occurrence) 
variety (L2_ENG) * animacy (human) 
variety (L2_JPN) * animacy (human) 
variety (L2_ENG) * animacy (inanimate) 
variety (L2_JPN) * animacy (inanimate) 

Note.  progressive level 'yes' coded as class 1. 

Estimate  SE 

z 

Odds 
Ratio 
 0.399   2.165    1.935   
0.772 
 1.302   5.076    1.248   
1.625 
-0.265   0.817   0.767    -0.324   
-0.414   0.378   0.661    -1.097   
 0.381   1.501    1.067   
0.406 
 0.349   1.187    0.491   
0.171 
-1.057   0.452   0.348    -2.337   
-0.804   0.351   0.448    -2.286   
-0.809   0.324   0.445    -2.496   
-0.660   0.343   0.517    -1.921   
 0.330   1.148    0.417   
0.138 
-1.137   0.805   0.321    -1.412   

Wald Test 

95% Confidence 
interval 

Wald 
Statistic 
3.742 
1.558 
0.105 
1.204 
1.140 
0.241 
5.461 
5.228 
6.229 
3.689 
0.174 

1.995 

df  p 
 1   0.053  
 1   0.212  
 1   0.746  
 1   0.273  
 1   0.286  
 1   0.623  
 1   0.019  
 1   0.022  
 1   0.013  
 1   0.055  
 1   0.677  
 1   0.158  

Lower 
bound 
-0.010 
-0.926 
-1.866 
-1.155 
-0.340 
-0.513 
-1.943 
-1.492 
-1.444 
-1.333 
-0.510 

Upper 
bound 
1.555 
4.176 
1.336 
0.326 
1.152 
0.855 
-0.170 
-0.115 
-0.174 
0.014 
0.785 

-2.714 

0.441 

-0.010   0.667   0.990    -0.015    2.338×10-4   1   0.988  
 1   0.366  
 1   0.199  
 1   0.081  
 1   0.120  
 1   0.070  

 0.713   1.905    0.903   
0.644 
 0.579   2.103    1.283   
0.743 
 0.719   3.513    1.746   
1.256 
 0.624   2.643    1.557   
0.972 
-2.036   1.123   0.131    -1.814   

0.816 
1.647 
3.050 
2.423 

3.289 

-1.318 

-0.753 
-0.392 
-0.154 
-0.252 

-4.236 

1.200 

 1.027   3.319    1.168   

1.365 

 1   0.243  

-0.813 

1.297 

2.042 
1.879 
2.666 
2.195 

0.164 

3.212 

-1.006   0.463   0.366    -2.174   

4.727 

 1   0.030  

-1.912 

-0.099 

-1.290   0.525   0.275    -2.459   

6.047 

 1   0.014  

-2.319 

-0.262 

1.253 

 0.597   3.502    2.101   

4.413 

 1   0.036  

0.084 

2.422 

 0.597   8.003    3.485   
2.080 
-2.406   1.224   0.090    -1.965   
-0.312   0.702   0.732    -0.444   
-0.928   1.238   0.395    -0.750   
-0.115   0.760   0.892    -0.151   

12.145 

3.860 
0.197 
0.563 
0.023 

 1  < .001  
 1   0.049  
 1   0.657  
 1   0.453  
 1   0.880  

0.910 

3.250 

-4.805 
-1.688 
-3.355 
-1.604 

-0.006 
1.064 
1.498 
1.375 

100 

 
 
  
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5.1. Addressing the research questions 

V. DISCUSSION 

This study, which is collostructional, frequency based, and multivariate, provides a view of the 

progressive construction -ko iss in Korean across varieties and textbooks. This study also shows 

the importance of considering constructions from multiple statistic perspectives (including 

collostructional analysis and regression modeling) and qualitative perspectives (exploring how a 

construction is introduced in textbooks in addition to quantifying the verbs with the construction) 

to develop a robust understanding of how a construction is used in terms of lemmas 

(dis)associated with it and to address linguistic factors that may influence the choice of one 

particular construction over another. These methodologies allow me to address three overarching 

research questions, which are (i) what are the distinctive collexemes for the progressive and non-

progressive across L1 and L2 written Korean varieties, (ii) what linguistic factors influence the 

choice of a progressive or a non-progressive, and (iii) how are Korean language textbooks 

incorporating the progressive -ko iss construction overall and across levels?  

Research question (i) addressed verbs and their preference for the progressive (or non-

progressive) in terms of association strengths measured in collostructional strength, specifically 

using the distinctive collexeme analysis method. Overall, the results are promising in that both 

learner varieties exhibited wide variety in their choice of verbs in the progressive in the writing, 

much of which fell in-line with the L1 corpus data. Of particular interest was the fact that L1 

English learners, despite the English language overall featuring fewer stative progressives than 

the Korean language does, key verbs which are highly associated with and well attested to co-

occur with the progressive -ko iss construction. Verbs such as sal (live), al (know), and neuggi 

(feel), stative verbs which appearing in the progressive in Korena, were found to be distinctive 

101 

 
 
collexemes in the L1 English learner data, showing a positive sign for uptake of typologically 

distinct forms from the learners’ L1. In terms of sheer number of distinctive collexemes, L1 

Japanese learners did overshadow the L1 English learners, including sal (live), gaji (have/hold), 

gidae (hope), mid (believe). The L1 Japanese learners using more progressives overall and more 

progressives that are particularly stative and mental verbs also found to be distinctive collexemes 

in the L1 data is not surprising as Japanese also allows for stative and mental verbs to take the 

progressive construction. This finding highlights the fact that potential entrenchment of the form 

(-ko iss progressive construction) with stative and mental verb readings can be a point of issue 

for learners of Korean whose L1s are typologically distant from Korean in their form-function 

mapping of progressive constructions and stative/mental meanings. 

With research question (ii), I employed logistic regression analysis to look at the big 

picture of what linguistic factors may influence the choice of a progressive across varieties, to 

varying degrees of success. I created three models, one without interaction effects (model 1), and 

two including interaction effects (model 2; model 3), which led me to modify the levels of 

semantic domain included in the final model 3. While model 3 has some weakness in terms of 

CIs including 1, as this study is exploratory, I will mention results from both model 1 and model 

3, with cautious optimism when discussing model 3.  

The first model showed that on their own, activity and stative verbs were more likely to 

predict the usage of the progressive -ko iss, whereas achievements showed a trend towards 

preferring the non-progressive. As some multifactorial studies on the progressive across varieties 

of English have shown that achievement verbs can trigger a progressive construction in academic 

writing (see, for example, Rautionaho et al., 2018) I anticipated that L1 English learners of 

Korean might tend towards using achievement verbs in the progressive in Korean. However, this 

102 

 
 
 
was not the case. Considering model 3 and the interactions found between variety and aktionsart 

and variety and semantic domain, results suggest learners may use the progressive -ko iss more 

commonly with occurrence verbs in their writing than L1 Korean speakers. Learners also appear 

to be using mental verbs less often in their writing, which may be in part due to typological 

differences. However, given that model 3 suffers from some issues with Confidence Intervals, 

more data must be collected in order to confirm this apparent trend.  

Considering the textbooks featured in this study leads to research question (iii), asking 

what verbs are most prevalent in the progressive in Korean across textbooks and stratified by 

level. Overall, the rate of usage of progressives was higher in KLEAR Integrated Korean than in 

New Sogang Korean. As textbook length increased, so too did the frequency of the use of the 

progressive -ko iss, though overall, the construction itself was still not used as frequently as 

expected. Diving into the verbs used in the progressive in the textbook corpora, the verbs used 

are largely action verbs (e.g., see, watch, listen, do). KLEAR Integrated Korean does incorporate 

a key stative verb, al (know), starting from the second level. New Sogang Korean incorporates al 

(know) as one of the top five most frequent verbs from level 4. Considering the results of the 

collostructional analysis and regression in combination with the relatively low frequencies of 

stative and mental verbs, it seems clear that future iterations of the textbooks could better 

represent real-world Korean language with more frequent inclusion of stative and mental verbs 

in the progressive, as well as using more diverse verbs in the progressive, considering distinctive 

collexemes that were absent in the learner data and not found frequently in the textbooks. It is 

also suggested that learners with L1 backgrounds that are both typologically similar and 

dissimilar to Korean could benefit from inclusion of such verbs, particularly as both learner 

groups have a trend to use mental verbs less frequently (calling back to the logistic regression).  

103 

 
 
My results support findings from, for example, Jang (2005), who found that the 

progressive -ko iss was largely one note and focused on the ‘action in progress’ meaning of -ko 

iss in the textbooks I surveyed. Further, in light of the combined textbook analysis in tandem 

with analysis of L1 and L2 data, in conversation with Kim (2014) who looked at -ko iss 

diachronically and noted its increasing usage with stative progressives (and overall similar rates 

of usage with both ‘action in progress’ and ‘stative/resultative’ meanings), I think it is time for 

textbooks to teach present -ko iss to learners as a construction with several distinct meanings and 

usages. That is to say, textbooks would better serve learners were they to introduce the ‘action in 

progress’ -ko iss at the lower levels as they do now, and then at the intermediate or advanced 

level re-introduce -ko iss and its multitude of usages, include with stative and mental verbs, its 

usage in writing to denote changes over time particularly when the subject is inanimate, and the 

breadth of verbs that can co-occur with the -ko iss construction that L1 English speakers might 

not anticipate. While textbooks are only one source of input, they can have an impact. Recall that 

Northbrook and Conklin (2019) found that even low-level learners were able to respond faster to 

a phrasal judgement task when the lexical bundles they encountered matched those in their 

textbooks. As the authors put it, this “indicates that… students are sensitive to whether items 

appeared in their books” and thus, “input given to students matters” (p. 828). Bearing this in 

mind, the argument to incorporate more usages of the -ko iss construction with a variety of verbs 

can have a positive impact on student uptake of the construction. Relating this back to 

Gabrielatos (2006), clearly, as input from textbooks is effective for students, informing their 

development using corpora can only aid students in their language learning. By revamping the 

representation of -ko iss, textbooks would be better prepared to serve learners and provide them 

with accurate descriptions and examples of this complex Korean construction.  

104 

 
 
5.2. Implications for the field of (Korean) second language acquisition 

This study also highlights the merit of using multiple statistical approaches to investigate 

large corpus data. Specifically, the results from the collostructional analysis, the regression, and 

the frequency analysis in the textbook section revealed how looking at the data from one singular 

method may cause overgeneralization of certain findings. Take, for example, the verb al (know). 

In the collostructional analysis, al was found to be a distinctive collexeme of the progressive in 

both the L1 Korean and L1 English L2 Korean varieties, but not in the L1 Japanese L2 Korean 

variety. This finding is initially surprising, as typological (dis)similarities across languages 

would suggest the Japanese learner language would be more likely to include mental, stative 

progressives such as al as distinctive collexemes. Further, the mental verb category was also 

found to cause a dispreference in the learner group. However, when considering the normalized 

frequencies of al across L1, L2, and textbook corpora, it became clear that in actuality al was 

used more (in terms of relative frequency) in the Japanese learner language than any other 

variety. Were this study to solely consider collostructional strengths or the results of the 

regression analysis the conclusion may have been that Japanese learner language lacks/underuses 

stative progressives and mental verbs with -ko iss. In fact, this triangulation approach suggests a 

more nuanced result, that while Japanese learners, overall, may be using mental verbs and stative 

progressives less than we see in the L1 data, when a verb functions similarly in Japanese (know 

in Japanese functioning similarly) they are using the verb at a higher frequency in particular with 

the construction, potentially due to transfer effects leading to entrenchment of that particular 

verb. Considering the fact that al was not a distinctive collexeme and that mental verbs generally 

disfavored the progressive in Japanese learner data, this shows that analysts considering 

interlanguage effects will need to consider, based on the languages in question, if any 

105 

 
 
particularly linguistic elements will require a deep dive beyond what association strength 

measures and inferential statistics can provide separately.  

5.2.1 Future directions 

In this section, I hope to highlight issues that arose during this project and provide insights and 

suggestions that may benefit future corpus-based projects on Korean. Namely, I will discuss 

modeling Korean using logistic regression. In sections 4.4 through 4.4.5, I conducted a logistic 

regression with the intention of determining which predictors may influence the choice of a 

progressive in L1 and L2 Korean. My annotation scheme was based on previous literature, and 

particularly, recent corpus studies which focus on English. However, what my analysis has 

highlighted is that some annotation schemes, particularly semantic domain, need adjusting in a 

follow-up study. For example, while semantic domains such as aspectual and causative were 

included in the present study, they were rarely found during annotation. In retrospect, the 

causative category could have been omitted considering Koreans typology. Causative verbs 

include allow or permit, however, in Korean, rather than a singular verb, another construction 

can be used to express this meaning. Therefore, it is perhaps not surprising that verbs with a 

causative reading were so sparse in the Korean data across all varieties. This highlights the 

importance of considering Korean’s distinct features when selecting annotations. A future study 

could eliminate aspectual and causative verbs from the annotation scheme altogether.  

Going forward, there is also room to improve the annotation schemes to tease out the 

nuances in the usage of key stative and mental verbs. In this study, many stative and mental 

verbs were found to be distinctive collexemes in L1 and L2 varieties of Korean, and usage of 

such verbs, though limited, appeared in textbooks as well, and a verb being categorized as a 

stative verb (aktionsart) made it more likely to trigger the use of a progressive according to 

106 

 
 
model 1 as well. Thus, to make the analysis of stative and mental verbs more robust, Korean-

specific aktionsart-esque categories can be introduced to a future multivariate analysis. For 

example, Lee (2006), in her paper on stative progressives in Korean takes the stance that ‘know-

type’ verbs, such as al (know) are punctual, and emotion verbs such as sarangha (love) are 

durative. Thus, a future analysis could consider whether punctuality or durativity of a stative 

verb in Korean lend themselves more to the progressive, and under what context. To add to that, 

some scholars have even suggested that verbs such as al (know) may even be accomplishments 

(Hong, 1991) depending on the event description that led someone to come to know something. 

Thus, a careful analysis of key stative and even mental verbs in Korean, where more language 

specific annotations are employed may help tease apart what makes Korean stative and mental 

verbs so unique in their usage in the -ko iss construction. Findings would have clear implications 

for pedagogy and materials development.  

5.3. Limitations 

There are several limitations to this study, some of which may serve as a guide for future studies 

and demonstrate the need for further development of learner corpora of languages other than 

English. First, a critical limitation of this study is in fact the L1 and L2 corpora that were 

available for this study. While both L1/L2 corpora were compiled and made available by the 

National Institute of Korean Language (NIKL) and feature relatively large amounts of data, they 

are not directly comparable with each other. For example, the NIKL L1 corpus is akin to the 

British National Corpus (BNC) as the written corpus comprises novels, short stories, 

newspapers, articles, opinion pieces, etc. On the other hand, the learner essays submitted to 

NIKL when compiling the corpus ranged in topic, including argumentative essays, opinion 

essays, and personal narrative essays. However, as these are some of the largest and most widely 

107 

 
available Korean language corpora, they were selected for this study. A prime example of the 

type of corpus that the field of Korean corpus linguistics is in need of is the International Corpus 

Network of Asian Learners of English (ICNALE; Ishikawa, 2023). The ICNALE corpus consists 

of spoken and written data contributed by L1 and L2 speakers of English, so that the language 

data is controlled for genre, and well documented for a speaker’s L1, proficiency, task, and other 

pertinent background information such as years studying English, all of which allows for robust 

interlanguage comparisons to be made. As of yet, such a corpus does not exist for the Korean 

language, and so the findings in the present study must be taken with the understanding that 

comparisons could shift should a more balanced corpus arise.  

In terms of the data analysis itself, due to (i) the massive amount of data to be extracted 

and (ii) time and resource constraints, this study only focused on exploring the prototypical 

action in progress -ko iss progressive construction in Korean. However, as noted in the literature, 

there are several ways to express continuous aspect in Korean, including other constructions 

(such as neun jung or a/eo iss) which constitute full examinations in papers of their own to create 

a full understanding of the continuous aspect in L1, L2, and textbook Korean. In terms of the 

learner data itself, while proficiency was originally intended to be considered as a factor in the 

regression analysis, unlike ICNALE, the NIKL learner data does not provide verified 

information on a learner’s language proficiency, such as a c-test or Test of Proficiency in Korean 

(TOPIK) score (ICNALE, in most cases, is able to provide c-test or TOEFL scores). Proficiency 

information in the NIKL learner corpus is based on level of Korean class (e.g., level 1, level 2), 

which as any classroom teacher can attest to, does not necessarily correspond to a learner’s 

actual language proficiency. In existing literature, some corpus linguists have found interactions 

between explanatory variables such as genre or tense, however, due to time constraints for data 

108 

 
  
extraction, cleaning, and manual annotation by two raters, tense was not considered in this 

analysis. A follow up study should consider tense as one explanatory variable in the choice of a 

(non)progressive across varieties of Korean, keeping in mind the large dataset that this will lead 

to and the amount of time necessary for manual annotation. Genre can also be considered in 

future studies, provided a corpus of Korean language is created to be both comparable across 

speaker varieties and documented so that the genre is known to the analyst. Finally, future 

studies may be able to incorporate fixed effects into their models, such as speaker, which was 

unable to be tested in this study due to a lack of speaker information for the L1 data. 

Finally, in terms of the textbook corpus, levels 1 through 4 of two of the commonly used 

textbook series for teaching Korean were included. Future studies could benefit from adding 

other textbook series to the corpus to compare the usage of the progressive across multiple 

volumes. Further, while levels 1 through 4 were included in the present study, in fact, both series 

have more advanced volumes (through 5 and 6). While these are not often used in teaching in the 

North American context due to each volume generally corresponding to an academic year of 

study, for Korean programs serving advanced learners who may use those advanced volumes, 

adding them to the analysis could provide useful information for textbook developers and 

language teachers. 

5.4. Pedagogical Implications 

5.4.1. For teachers 

In terms of pedagogical implications, I will discuss them here in terms of implications for 

teachers of Korean and implications for language materials and textbook developers. First of all, 

it is clear that learner language differs from L1 language in terms of the verbs that are associated 

with the progressive, as well as the wide variety of semantic domains that those verbs can fall 

109 

 
 
into. Functionally, learners are limited in terms of their ability to use stative and mental verbs 

with the progressive in their writing. As such verbs also appear commonly in authentic written 

texts it is important for teachers to at the very least make learners aware of this form-function 

connection in the classroom. To facilitate this, genres which learners are interested in, such as 

manhwa (Korean comics) or clips from Korean shows can be used in lower levels as they will 

include examples of stative mental verbs in the progressive. Likewise, higher level learners can 

be exposed to news articles or short stories and novels, and teachers can modify the text 

complexity to accommodate their learners while maintaining examples including the progressive. 

Additionally, highlighting the progressive form with stative and mental verbs in class through 

discussion where learners are required to reproduce the form can help facilitate practice and 

uptake.  

Empowering learners with authentic materials in the classroom has been found to 

motivate learners at all levels (Bahrani et al., 2014). As learners may be demotivated if the texts 

are too difficult (Sample, 2015), it is important for teachers to modify authentic materials for 

intermediate or emerging advanced learners of Korean. One actionable recommendation is for 

teachers to start by using authentic news articles about topics learners are familiar with. For 

example, news sites such as Huffington Post Korea often publish articles on topics learners are 

interested in and familiar with, including Korean pop culture but also extending to celebrities and 

headlines trending outside of Korea. Teachers can use such articles as a gateway to authentic 

materials while avoiding issues of topic unfamiliarity.  

Additionally, the collostructional analysis revealed that certain distinctive collexemes in 

the learner data may be being overused when compared with the L1 data. For example, sal (live) 

and manhaji (increase) were distinctive collexemes in learner data. However, in the L1 written 

110 

 
corpus data, equivalent but more academic terms, namely geojuha (live/reside) and jeunggaha 

(increase) were identified as distinctive collexemes. Learners may rely on simpler terms they 

have learned early on, and thus these terms are well entrenched in their mental lexicons. Thus, 

when incorporating authentic materials, Korean language teachers can help learners improve 

their Korean writing to appear more academic by  making these equivalent verb forms salient 

and helping learners identify when to use each verb type.  

5.4.2. For textbook developers and materials designers 

For textbook developers, a major implication is that the use of the progressive needs to be 

more widespread in the textbooks, particularly in terms of mental verb representation as the 

analysis has found that both learner groups used mental verbs in their writing significantly less 

than the L1 Korean corpus, and the variety of verbs that learners associated with the progressive 

as a whole was significantly lower. As textbooks are one main source of input for learners, 

including such examples at all levels is critical. Lower-level textbooks can incorporate 

stative/mental verbs in the progressive to dialogues as they are used in spoken language, which 

learners will practice in the classroom, thus sowing the seeds for them to gain awareness of the 

form-function mapping and be more inclined to notice and acquire progressives when they 

appear in texts at the higher level. This is of particular note for textbook series which altogether 

lack examples of the progressive in the lower-level volumes as was found in this study.  

In addition to incorporating more stative and mental verbs even at lower levels, textbooks 

could aid learners’ uptake of the usage of the progressive by incorporating readings which 

include inanimate subjects of the progressive verb. The logistic regression analysis here revealed 

that, in L1 Korean writing, the progressive was more likely to be used when the subject was 

inanimate, and this was commonly seen in the L1 corpus. Further, while it was anticipated that 

111 

 
 
 
achievement verbs (which express punctual events) would be used in the progressive by L1 

English speakers, this was not borne out in the results. In fact, L1 English speakers were far less 

likely to use an achievement verb with the progressive than was seen in the L1 corpus data. So, 

at the very least, incorporating readings in the texts which include achievement verbs that were 

associated with the progressive in the textbooks (e.g., natana - appear, geuchi - stop/cease; 

ireugi when ireugi is used with the semantic meaning of trigger).  

Further, for textbooks specifically, incorporating grammar explanations of the variety of 

uses the progressive can have would benefit learners. The textbook series presented in this study, 

when they introduce the progressive, include grammar explanations as to how the progressive -

ko iss describes an action in progress. For example, qualitative exploration of the textbooks 

revealed that when -ko iss was explicitly taught it was used with actions such as watch, listen, 

wash, clean, and so forth. Mental and stative verbs were not represented in the explicit teaching 

sections of the texts, and only appeared later on in the textbook series incidentally. In fact, it 

appears that textbooks, particularly at the lower levels, incorporate more instances of the 

progressive being used in tandem with the present to show learners how its usage is option (e.g., 

asking what are you doing with the main verb do in the simple present, and then responding I am 

drinking tea in the present progressive). While such distinctions are important, including 

grammar descriptions and examples with explanations of the progressive used with stative verbs 

and mental verbs in particular can help learners notice and acquire the forms.  

To strengthen the linguistic description of the -ko iss construction in textbooks, I 

recommend introducing it at least twice at different levels. In the beginner levels, introducing -ko 

iss as ‘action in progress’ can help facilitate the acquisition of this prototypical form-function 

mapping that learners can easily practice in the classroom. At more advanced levels, 

112 

 
 
 
reintroducing the progressive as it is used with various semantic senses in both spoken and 

written language can also be beneficial and allow learners, particularly L1 English speakers, to 

notice the forms of the progressive which are less common in English. Namely, this amounts to 

teaching frequently taught chunks such as al (know), gaji (have/hold), jeunggaha (increase), 

bododwe (be reported) among others which were found to be distinctive of the progressive in L1 

writing to learners. At the very least, a re-examination of the -ko iss construction and its various 

semantic meanings beyond simply ‘action in progress’ is warranted and would be beneficial for 

learners in their Korean language learning. 

113 

 
 
 
REFERENCES 

Abbot, K., & Tomasello, M. (2006). Exemplar-learning and schematization in a usage-based 
account of syntactic acquisition. The Linguistic Review, 23(3), 275-290. doi: 
10.1515/TLR.2006.011 

Ahn, Y. (1995). The aspectual and temporal system of Korean: From the perspective of the two-
component theory of aspect. Unpublished Doctoral Dissertation, University of Texas at Austin. 

Andersen, R. W. (1990). Models, processes, principles and strategies: Second language 
acquisition inside and outside the classroom. In B. VanPatten & J. F. Lee (Eds.), Second 
language acquisition-Foreign language learning, Multilingual Matters, 45-78. 

Andersen, R. W. (1991). Developmental sequences: The emergence of aspect marking in second 
language acquisition. In T. Huebner & C. A. Ferguson (Eds.), Tense-aspect morphology in L2 
acquisition, 79-105. John Benjamins. 

Andersen, R. W., & Shirai, Y. (1994). Discourse motivations for some cognitive acquisition 
principles. Studies in Second Language Acquisition, 16, 133-156. 

Anderwald, L. (2012). “I’m loving it” – marketing ploy or language change in progress?. 
Presented at the Symposium: The pragmatics of aspect in varieties of English. 
https://doi.org/10.1080/00393274.2016.1208536  

Anthony, L. (2023). AntConc (Version 4.2.4) [Computer Software]. Tokyo, Japan: Waseda 
University. Available from https://www.laurenceanthony.net/software 

Anthony, L. (2022). TagAnt (Version 2.0.5) [Computer Software]. Tokyo, Japan: Waseda 
University. Available from https://www.laurenceanthony.net/software 

Bardovi-Harlig, K., & Comajoan-Colomé, L. (2020). The aspect hypothesis and the acqusition of 
L2 past morphology in the last 20 years: A state-of-the-scholarship review. Studies in Second 
Language Acquisition, 42, 1137-1167. doi:10.1017/S0272263120000194 

Bahrani, T., Tam, S. S., & Zuraidah, M. D. (2014). Authentic Language Input Through 
Audiovisual Technology and Second Language Acquisition. Sage Open, 4(3), 1-8. 
doi:10.1177/2158244014550611 

Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear mixed-effects models 
using lme4. Journal of Statistical Software, 67, 1-48. 

Belli, S. A. (2018). An analysis of stative verbs used with the progressive aspect in corpus-
informed textbooks. English Language Teaching, 11(1), 120-135. doi: 10.5539/elt.v11n1p120 

Biber, D. (1999). Longman Grammar of Spoken and Written English. Longman. 

114 

 
 
 
 
 
 
 
 
 
 
 
 
 
Biber, D., Johansson, S., Leech, J., Conrad, S., & Finegan, E. (2021). Grammar of Spoken and 
Written English. Amsterdam & Philadelphia: John Benjamins Publishing Company. 

Biber, D., & Conrad, S. (2010). Corpus linguistics and grammar teaching. Available at: 
www.longmanhomeusa.com/content/pl_biber_conrad_monograph_lo_3.pdf 

Brown, L., & Yeon, J. (2010). Experimental research into the phases of acquisition of Korean 
tense-aspect: Focusing on the progressive marker “-ko issta.” Journal of Korean Language 
Education, 21(1), 151–173.  

Bybee, J. L. (2013). Usage based theory and exemplar representations of constructions. In T. 
Hoffmann & G. Trousdale (Eds.), The Oxford Handbook of Construction Grammar (pp. 49-69). 
The Oxford Handbook of Construction Grammar (2013; online edn, Oxford Academic, 16 Dec. 
2013), https://doi.org/10.1093/oxfordhb/9780195396683.013.0004 

Chae, H-R. (2018). The pseudo-resultative {V-ko (iss)} Construction in Korean. Language 
Research, 54(2), 157-200. https://doi.org/10.30961/lr.2018.54.2.157 

Davies, M. (2008-) The Corpus of Contemporary American English (COCA). Available online at 
https://www.english-corpora.org/coca/. 

Deshors, S. C., (2011). A multifactorial study of the uses of may and can in French-English 
interlanguage. A University of Sussex DPhil theses, Available online via Sussex Research 
Online: https://core.ac.uk/download/pdf/2710234.pdf 

Deshors, S. C., & Gries, S. T. (2014). A case for the multifactorial assessment of learner 
language: The uses of may and can in French-English interlanguage. In D. Glynn and J. A. 
Robinson (Eds.). Corpus Methods for Semantics: Quantitative studies in polysemy and 
synonymy. John Benjamins Publishing Company. 

Deshors, S. C., & Gries, S. T. (2023). Using corpora in research on second language 
psycholinguistics. In A. Godfroid & H. Hopp (Eds.). The Routledge Handbook of Second 
Language Acquisition. Routledge. 

Flowerdew, L. (1998). Corpus linguistic techniques applied to textlinguistics. System, 26(4), 
541-552. https://doi.org/10.1016/S0346-251X(98)00039-6 

Fokkema, M., Smits, N., Zeileis, A., Hothorn, T., & Kelderman, H. (2018). Detecting treatment-
subgroup interactions in clustered data with generalized linear mixed-effects model trees. 
Behavior Research Methods, 50, 2016-2034. doi: https://doi.org/10.3758/s13428-017-0971-x 

Freund, N. (2016). Recent change in the use of stative verbs in the progressive form in British 
English: I'm loving it. Language Studies Working Papers, 7, 50-61. 

115 

 
 
 
 
 
 
 
 
 
 
 
 
 
Fuchs, R., & Werner, V. (2018). The use of stative progressives by school-age learners of 
English and the importance of the variable context. International Journal of Learner Corpus 
Research, 4(2), 195-224. https://doi.org/10.1075/ijlcr.00004.int 

Gabrielatos, C. (2005). Corpora and language teaching: Just a fling or wedding bells? The 
Electronic Journal for English as a Second Language, 8(4).  

Granath, S. & Wherrity, M. (2014). "I'm loving you - and knowing it too": Aspect and so-called 
stative verbs. Rhesis: Linguistics and Philology, 4(1), 2-22. 
http://urn.kb.se/resolve?urn=urn:nbn:se:kau:diva-31699 

Granger, S. (2009). The contribution of learner corpora to second language acquisition and 
foreign language teaching: A critical evaluation. In K. Aijmer (Ed.), Corpora and Language 
Teaching. John Benjamins Publishing Company. Permalink: 
http://digital.casalini.it/9789027289988 

Gries, S. T., & Stefanowitsch, A. (2004). Extending collostructional analysis: A corpus-based 
perspective on 'alternations'. International Journal of Corpus Linguistics, 9, 97-129. 
https://doi.org/10.1075/ijcl.9.1.06gri 

Gries, S., Hampe, B. & Schönefeld, D. (2005). Converging evidence: Bringing together 
experimental and corpus data on the association of verbs and constructions. Cognitive 
Linguistics, 16(4), 635-676. https://doi.org/10.1515/cogl.2005.16.4.635 

Gries, S. T. (2014). Coll.analysis 3.5: A script for R to compute perform collostructional 
analyses. 

Gries, S. T., & Deshors, S. C. (2014). Using regressions to explore deviations between corpus 
data and a standard/target: Two suggestions. Corpora, 9(1), 109-136. DOI: 
10.3366/cor.2014.0053 

Gries, S. T. (2015). The most underused statistical methods in corpus linguistics: Multi-level 
(and mixed effects) models. Corpora, 10(1), 95-125. DOI: 10.3366/cor.2015.0068 

Hong, K-S. (1991). Argument selection and case marking in Korean. Unpublished Doctoral 
Dissertation, Stanford University.  

Hothorn, T., & Zeileis, A. (2015). partykit: A modular toolkit for recursive partytioning in R. 
Journal of Machine Learning Research, 16, 3905-3909. Available from 
https://jmlr.org/papers/v16/hothorn15a.html 

Hundt, M., & Vogel, K. (2011). Overuse of the progressive in ESL and learner Englishes - fact 
or fiction?. Studies in Corpus Linguistics, 44, 145-166. https://doi.org/10.1075/scl.44.08vog 

Hundt, M., Rautionaho, P., & Strobl, C. (2020). Progressive or simple? A corpus-based study of 
aspect in World Englishes. Corpora, 15(1), 77-106–106. doi: 10.3366/cor.2020.0186 

116 

 
 
 
 
 
 
 
 
 
 
 
 
 
Jang, M.-s. (2005). The improved plans for teaching Korean tense and aspect forms: Focusing on 
hanta type and hako issta type. Journal of Korean Language Education, 16(3), 305–330.  

Jeong, S. J. (2011). A study on the description methods of the adverbial case postpositions for 
Korean education based on cognitive linguistics. The Korean Language and Literature, 112, 79–
110.  

Jung, B. K. (2022). The nature of L2 input: Analysis of textbooks for learners of Korean as a 
second language. Korean Linguistics, 18(2), 182-208. https://doi.org/10.1075/kl.20001.jun 

Kim, H., Kang, B., & Hong, J. (2007). 21st Century Sejong Corpora (to be) completed. The 
Korean Language in America, 12, 31-42. JSTOR, http://www.jstor.org/stable/42922169. 

Kim, S. K. (2011). Education method for the adverb postpositions of ‘ey’, ‘eyse’, ‘lo’ in the 
Korean language. Kwukhakyenkwulonchong, 8, 199–236.  

Kim, Y., & Guo, J. (2016). A study on the acquisition of Korean adverbial case marker ey in 
spoken production by Chinese Korean L2 learners. Korean Education Research, 38, 1-26. 

Koprowski, M. (2005). Investigating the usefulness of lexical phrases in contemporary 
coursebooks. ELT Journal, 59 (4), 322–32.  

Kranich, S. (2010) Progressive in modern English: A corpus-based study of grammaticalization 
and related changes. Amsterdam: Rodopi. doi: 10.1163/9789042031449 

Lam, P. W. Y. (2009). Discourse particles in corpus data and textbooks: The case of Well. 
Applied Linguistics, 31(2), 260-281. https://doi.org/10.1093/applin/amp026 

Lee, E. (2006). Stative progressives in Korean and English. Journal of Pragmatics, 38, 695-717. 
doi:10.1016/j.pragma.2005.09.006  

Northbrook, J., & Conklin, K. (2019). Is what you put in what you get out?: Textbook-derived 
lexical bundle processing in beginner English learners. Applied Linguistics, 40(50), 816-833. 
doi:10.1093/applin/amy027  

Rautionaho, P. (2014). Variation in the progressive: A corpus-based study into World Englishes. 
Tampere: Tampere University Press. 

Tomasello, M. (2003). Constructing a language: A usage-based theory of language acquisition. 
Harvard University Press. 

Rautionaho, P., & Deshors, S. C. (2018). Progressive or not progressive?: Modeling the 
constructional choices of EFL and ESL writers. International Journal of Learner Corpus 
Research, 4(2), 225-252. https://doi.org/10.1075/ijlcr.16019.rau 

117 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Rautionaho, P. (2020). Revisiting the myth of stative progressives in world Englishes. World 
Englishes, 41, 183-206. DOI: 10.1111/weng.12520 

Römer, U. (2004). Comparing real and ideal learning input: The use of an EFL textbook corpus 
in corpus linguistics and language teaching. In G. Aston, S. Bernardini, & D. Stewart (Eds.). 
Corpora and Language Learners. John Benjamins Publishing Company. 
https://doi.org/10.1075/scl.17.12rom 

Römer, U. (2005). Progressives, Patterns, Pedagogy: A corpus-driven approach to English 
progressive forms, functions, contexts and didactics. John Benjamins Publishing Company. doi: 
https://doi.org/10.1075/scl.18 

Römer, U. (2006). Where the computer meets language, literature, and pedagogy: Corpus 
analysis in English studies. In A. Gerbig, A. Müller-Wood (Eds.). How Globalization Affects the 
Teaching of English: Studying Culture Through Texts. Lampeter: E. Mellen Press. 81-109. 

Römer, U. (2011). Corpus research applications in second language teaching. Annual Review 
of Applied Linguistics, 31, 205-225. doi: 10.1017/S0267190511000055 

Salaberry, R., & Shirai, Y. (2002). L2 acquisition of tense-aspect morphology. In Salaberry, R,. 
& Shirai, Y. (Eds.). L2 Acquisition of Tense-Aspect Morphology. John Benjamins Publishing 
Company. 

Schmidt, R. (1990). The role of consciousness in second language learning. Applied Linguistics, 
11(2), 129-158.  

Sinclair, J. M. (Ed.) (1987). Looking up: An account of the COBUILD project in lexical 
computing. London: Collins ELT. 

Sinclair, J. (1997). Corpus evidence in language description. In A. Wichmann, S. Fligelstone, T. 
McEnery, & G. Knowles (Eds.), Teaching and Language Corpora, 27-39. London: Longman.  

Sinclair, J. M. (Ed.) (2004). How To Use Corpora in Language Teaching. Amsterdam and 
Philadelphia: John Benjamins Publishing Company. 

Stefanowitsch, A., & Gries, S. T. (2003). Collostructions: Investigating the interaction of words 
and constructions. International Journal of Corpus Linguistics, 8, 209-243. 
https://doi.org/10.1075/ijcl.8.2.03ste 

Straka, M., Hajič, J., & Straková, J. (2016). UDPipe: Trainable pipeline for processing CoNLL-
U files performing tokenization, morphological analysis, POS tagging and parsing. In 
Proceedings of the Tenth International Conference on Language Resources and Evaluation 
(LREC 2016), Portorož, Slovenia, May 2016. 

Timmis, I. Corpora and materials: Towards a working relationship. In B. Tomlinson (Ed.). 
Developing materials for language teaching. Bloomsbury Publishing. 

118 

 
 
 
 
 
 
 
 
 
 
 
 
 
Vendler, Z. (1957). Verbs and Times. The Philosophical Review, 66, 143-160. 
https://doi.org/10.2307/2182371  

Virtanen, T. (1996). The progressive in NS and NNS student compositions: Evidence from the 
International Corpus of Learner English. In M. Ljung (Ed.), Corpus-based Studies in English 
Papers from the Seventeenth International Conference on English Language Research and 
Computerized Corpora, Rodopi B.V: Amsterdam. 

Yeon, J., & Brown, L. (2011). Korean, A Comprehensive Grammar. New York: Routledge.  

Zaenen, A., Carletta, J., Garretson, G., Bresnan, J., Koontz-Garboden, A., Nikitina, T., 
O'Connor, M. C., & Wasow, T. (2004). Animacy encoding in English: why and how. In 
Proceedings of the ACL-04 Workshop on Discourse Annotation. Available at: 
https://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.154.7 

119 

 
 
 
 
 
 
 
APPENDIX A: DISTINCTIVE COLLEXEME ANALYSIS I 

Table A-1.  

Distinctive collexemes for the (non)progressive in L1 written Korean data 

Progressive 

Coll.strength 

Non-progressive 

Coll.strength 

bo – ‘see’ 

beoli – ‘start/begin’ 

bad – ‘receive’ 

balghi – 

‘light/brigthen’ 

gyeogg – 

‘experience (esp. 

hardship)’ 

banbalha – 

‘oppose’ 

du – ‘put/set/place’ 

jarijab – 

‘settle/situate’ 

nopaji – ‘rise’ 

naenoh – ‘put/take 

out’ 

sal – ‘live’ 

jaegidwe – ‘be 

raised/made’ 

319.52 

148.47 

116.98 

114.51 

moreu – ‘not know’ 

boi – ‘be seen’ 

malha – ‘speak’ 

saenggagha – ‘think’ 

727.00 

407.30 

394.15 

195.83 

104.03 

na – ‘happen’ 

160.38 

102.15 

sijagha – ‘start’ 

107.96 

84.31 

78.29 

77.46 

77.02 

74.97 

74.71 

deuleoga – ‘go in’ 

yeolli – ‘open’ 

ireu – ‘reach/get to’ 

sihaengdwe – ‘go into 

effect’ 

ju – ‘give’ 

pyeolcyeoji – ‘spread’ 

96.15 

92.38 

90.95 

78.30 

76.51 

63.27 

isddareu – ‘occur in 

73.95 

jumogdwe – ‘be 

61.56 

succession’ 

hwagsandwe – 

‘spread’ 

watched’ 

71.54 

deulli – ‘be heard’ 

54.97 

120 

 
 
 
 
Table A-1 (cont’d). 

nao – ‘come out’ 

70.77 

ggobhi – ‘be in a 

53.20 

gaji – ‘have/hold’ 

gajchu – 

‘prepare/be 

equipped’ 

al – ‘know’ 

69.49 

65.19 

range’ 

geolli – ‘take time’ 

bara – ‘hope’ 

45.49 

45.27 

64.84 

bulli – ‘be 

41.44 

referred/called as’ 

bij – ‘come into/be 

63.14 

manna – ‘meet’ 

41.20 

in 

conflict/criticism’ 

allyeoji – ‘be 

known’ 

alh – ‘suffer’ 

girogha – 

‘record/document’ 

sa – ‘buy’ 

dalli – ‘run’ 

eod – ‘gain’ 

geomtoha – 

‘review/examine’ 

gaj – ‘have/hold’ 

60.70 

deuleoo – ‘come in’ 

37.52 

58.20 

51.03 

51.03 

44.30 

44.16 

41.53 

saenggi – ‘form’ 

bureu – ‘call’ 

neom – ‘over/excess’ 

ollaga – ‘go up’ 

gidaedwe – ‘expect’ 

yeol – ‘open’ 

37.41 

36.82 

33.57 

31.42 

30.52 

27.05 

39.21 

jeonmangdwe – 

26.50 

‘view/predict’ 

beoleoji – ‘happen, 

37.88 

jonjaeha – ‘exist’ 

25.78 

take place’ 

naedabo – ‘predict’ 

sseu – ‘use’ 

37.42 

36.86 

jijeogha – ‘indicate’ 

dalha – ‘reach (e.g., 

25.40 

25.12 

level)’ 

moeu – ‘gather' 

36.22 

gweonha – ‘advise’ 

24.96 

121 

 
 
22.59 

22.03 

20.65 

15.56 

15.17 

15.06 

14.83 

14.21 

12.91 

Table A-1 (cont’d). 

sseu – ‘write’ 

35.98 

pulidwe – ‘be 

gidari – ‘wait’ 

gaha – ‘apply, spur, 

34.93 

32.68 

cause’ 

jeonha – ‘tell, 

convey, pass on 

information’ 

nuri – ‘enjoy’ 

explained’ 

jujangha – ‘assert’ 

dojeonha – 

‘challenge’ 

32.68 

bunseogdwe – 

18.75 

‘analyze’ 

32.28 

gyeoljeongha – 

17.62 

‘decide’ 

molli – ‘be driven 

32.07 

yeogyeoji – ‘be 

15.71 

to/into’ 

olli – ‘raise’ 

nori – ‘seek, aim’ 

32.07 

31.04 

considered as’ 

ddeona – ‘depart’ 

deungjangha – 

‘appear’ 

pyeolchi – ‘spread’ 

30.55 

haeseogdwe – 

naeri – ‘get off’ 

gojodwe – ‘tone up, 

enhance’ 

geuchi – ‘stop’ 

chujinha – ‘push 

ahead with sth, 

promote’ 

29.41 

28.88 

‘interpret’ 

yogudwe – ‘request’ 

johaha – ‘like’ 

27.18 

balgyeonha – 

‘discover’ 

26.34 

yeongyeoldwe – 

12.85 

‘connect’ 

boyuha – ‘posses’ 

26.17 

seolmyeongha – 

12.56 

jeonhaeji – ‘be 

25.93 

passed 

along/conveyed’ 

‘explain’ 

gangjoha – 

‘emphasize’ 

12.44 

122 

 
Table A-1 (cont’d). 

bul – ‘blow’ 

25.49 

uryeodwe – ‘be 

12.04 

25.49 

24.48 

concerned’ 

ggaedad – ‘realize’ 

gubundwe – ‘sort’ 

12.03 

11.91 

24.31 

mandeul – ‘make’ 

11.18 

ta - ‘ride’ 

ddeooreu – ‘rise, 

come up’ 

uryeoha – ‘be 

concerned or 

fearful’ 

hwaldongha – ‘do 

24.26 

deud – ‘listen’ 

10.74 

an activity’ 

cuiha – ‘be drunk 

or enraptured in 

something’ 

ginjangha – ‘worry’ 

junbiha – ‘prepare’ 

sam – ‘be 

considered as’ 

jiki – ‘protect’ 

eosgalli – ‘have a 

disagreement’  

beonji – ‘spread’ 

ssod – ‘spill, pour’ 

ssodaji – ‘pour, 

gush’ 

nah – ‘produce, 

spawn, give birth' 

ga – ‘go’ 

23.81 

mud – ‘ask’ 

10.49 

23.81 

23.71 

ggeutna – ‘end’ 

seonboi – 

‘show/present’ 

10.32 

9.70 

23.05 

chujeongdwe – ‘trace’ 

9.66 

22.13 

21.62 

jinaga – ‘pass’ 

neomchi – ‘overflow’ 

9.33 

9.14 

21.37 

yeosboi – ‘get a sense 

9.14 

21.37 

21.37 

of’ 

dolao – ‘return’ 

heureu – ‘flow’ 

20.76 

salpyeobo – 

‘examine/check’ 

19.99 

ddeu – ‘scoop’ 

8.86 

8.14 

8.10 

7.93 

123 

 
 
 
 
 
Table A-1 (cont’d). 

namgi – ‘save, set 

19.79 

jeulgi – ‘enjoy’ 

7.79 

aside sth’ 

beoti – ‘endure’ 

18.80 

salpi – ‘look (as in see 

7.79 

umjigi – ‘move’ 

maej – ‘bear, sign, 

enter into contract’ 

bododwe – ‘be 

reported’ 

yeogi – ‘regard as’ 

keoji – ‘get bigger’ 

mosaegha – ‘seek, 

find’ 

18.43 

18.20 

about something)’ 

ggichi – ‘influence’ 

chamgaha – ‘attend’ 

7.79 

7.01 

17.15 

balgyeondwe – ‘be 

6.82 

17.15 

16.72 

16.16 

discovered’ 

ja – ‘sleep’ 

nureu – ‘push’ 

sarangha – ‘love’ 

6.45 

6.45 

6.45 

injeongbad – ‘be 

15.51 

hwaginha – ‘confirm’ 

6.31 

recognized’ 

saenghwalha – 

‘live, as in make a 

living or live your 

life’ 

deonji – ‘throw’ 

geuri – ‘draw’ 

deureonae – 

‘expose’ 

pal – ‘sell’ 

yoguha – ‘request’ 

dolli – ‘turn’ 

heundeulli – 

‘shake’ 

15.14 

gobaegha – ‘confess’ 

5.59 

14.75 

14.74 

sui – ‘rest’ 

yeongeobha – ‘do 

business’ 

5.58 

5.58 

14.57 

jeogyongdwe – ‘get 

5.47 

14.23 

14.20 

13.82 

13.82 

used to’ 

nolla – ‘be surprised’ 

pyeonggadwe – ‘be 

rated’ 

ddareu – ‘follow’ 

gongyeonha – 

‘perform’ 

5.44 

5.25 

5.19 

4.73 

124 

 
Table A-1 (cont’d). 

saenggyeona – 

‘emerge, occur’ 

13.04 

gieogdwe – ‘be 

4.69 

remembered’ 

yaegoha – ‘notify 

13.04 

heoyongdwe – ‘be 

4.69 

previously/in 

advance’ 

permitted’ 

nopi – ‘increase’ 

12.87 

punggi – ‘give off 

4.69 

jibjungdwe – ‘be 

12.81 

focused’ 

smell’ 

bumbi – ‘be 

overcrowded’ 

4.56 

gyesogdwe – ‘be 

12.72 

jindanha – ‘diagnose’ 

4.56 

continued’ 

chajiha – ‘possess, 

12.70 

pyeonggaha – ‘rate’ 

4.42 

or take possession’ 

pyeonggabad – 

12.33 

jinae – ‘spend/pass’ 

4.33 

‘receive a ranking’ 

geumjiha – ‘be 

12.27 

tujaha – ‘invest’ 

prohibited’ 

chamyeoha – 

‘attend’ 

11.27 

meog – ‘eat’ 

bulanhaeha – ‘feel 

10.68 

gusaha – ‘have 

uneasy’ 

pum – ‘brood’ 

o – ‘come’ 

yujiha – ‘keep, 

maintain’ 

badadeulyeoji – 

‘accept something’ 

palli – ‘be sold’ 

10.68 

10.18 

10.18 

9.97 

9.64 

command of’ 

kyeo – ‘turn on’ 

yeogseolha – 

‘emphasize’ 

noneuiha – 

‘discuss/debate’ 

gamjidwe – 

‘sense/detect’ 

chusandwe – ‘be 

estimated’ 

125 

4.22 

3.95 

3.91 

3.91 

3.76 

3.57 

3.26 

3.13 

 
Table A-1 (cont’d). 

eongeubha – 

‘mention’ 

insigha – ‘be 

aware’ 

myosahah – 

‘describe’ 

ddi – ‘assume (as in 

take sth on)’ 

bultaeu – ‘burn’ 

josaha – 

‘investigate’ 

oichi – ‘shout’ 

simhoadwe – 

‘deepen’ 

chusanha – 

‘estimate’ 

bandaeha – 

‘oppose’ 

neuggi – ‘feel’ 

neuleona – 

‘increase’ 

baeu – ‘learn’ 

busangha – ‘float, 

emerge’ 

chaetaegha – 

‘choose/adopt (as in 

a resolution etc.)’ 

jibaeha – ‘rule, 

dominate’ 

9.21 

9.21 

9.21 

9.15 

9.11 

9.11 

9.11 

9.11 

8.95 

8.57 

8.34 

8.20 

7.80 

7.80 

7.80 

jarangha – ‘brag’ 

dwe – ‘become’ 

3.11 

3.10 

jihyangha – ‘pursue’ 

2.95 

bunryudwe – 

‘classify’ 

ihaeha – ‘understand’ 

balpyoha – ‘present’ 

2.70 

2.70 

2.63 

gwancheugdwe – ‘be 

2.60 

observed/predicted’ 

naemil – ‘stick/hold 

2.60 

out’ 

olmgi – ‘move’ 

2.60 

jaesiha – ‘suggest’ 

2.54 

weonha – ‘want’ 

gongyuha – ‘share’ 

neoh – ‘put in’ 

seo – ‘stand’ 

2.37 

2.35 

2.35 

2.35 

seoneonha – ‘declare’ 

2.35 

7.80 

teu – ‘open’ 

2.35 

126 

 
Table A-1 (cont’d). 

chisos – ‘rise, soar, 

surge’ 

dabbyeonha – 

‘reply’ 

dwechaj – ‘take 

back’ 

ginjangsiki – ‘make 

nervous’ 

ilgwanha – ‘be 

consistent in doing 

something’  

jigmyeonha – 

‘encounter’ 

taeu – ‘burn, singe’ 

yaecheugha – 

‘predict’ 

ilha – ‘work’ 

bonae – ‘send’ 

gominha – ‘worry’ 

jab – ‘grab’ 

jibjungha – ‘focus’ 

nanu – ‘distribute’ 

seonjeonha – 

‘propogate’ 

ddeoleoji – ‘fall, 

decrease’ 

7.57 

7.57 

7.57 

7.57 

7.57 

7.57 

7.57 

7.57 

7.52 

7.46 

7.28 

6.52 

6.43 

6.43 

6.43 

6.16 

iyongha – ‘use’ 

2.27 

injeongha – ‘accept’ 

2.24 

dolaga – ‘go back’ 

2.21 

gieogha – ‘remember’ 

2.19 

deuleoseo – ‘enter in’ 

2.19 

jaeanha – ‘offer’ 

bunpoha – ‘distribute’ 

ilh – ‘lose/be 

deprived’ 

apseo – ‘get head’ 

ihaedwe – ‘be 

understood’ 

naga – ‘go out’ 

sidoha – ‘try’ 

banghwangha – 

‘wander’ 

chugadwe – ‘be 

added’ 

gajyeoga – ‘take’ 

galli – ‘be 

changed/divided’ 

1.95 

1.76 

1.76 

1.76 

1.76 

1.69 

1.66 

1.63 

1.63 

1.63 

1.63 

127 

 
 
 
 
 
Table A-1 (cont’d). 

daedudwe – ‘come 

6.06 

ggojib – ‘pinch’ 

1.63 

bbae – ‘subtract’ 

gueonyuha – ‘invite’ 

heoyongha – ‘permit’ 

goreu – ‘choose’ 

1.39 

1.39 

1.39 

1.32 

to the fore, be on 

the rise’ 

ddeolchi – ‘shake 

off, ride oneself of’ 

meomureu – ‘stay’ 

pyoha – ‘express’ 

uihyeobha – 

‘intimidate’ 

jagyongha – ‘act, 

function’ 

unyeongha – 

‘manage (e.g., 

business)’ 

gonggeubha – 

‘supply, provide’ 

gyesogha – 

‘continue’ 

buri – ‘manage, 

handle’ 

geol – ‘count on 

hopes or 

expectations’ 

alli – ‘tell’ 

dwepuliha – 

‘repeat’ 

salaga – ‘live’ 

sihaengha – ‘carry 

out, enforce’ 

6.06 

6.06 

6.06 

6.06 

6.01 

5.99 

5.78 

5.56 

5.45 

5.22 

5.11 

5.11 

5.11 

5.11 

128 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d). 

sseu_singyeong – 

5.11 

‘care about 

something’ 

gareuchi – ‘teach’ 

sidalli – ‘suffer 

from something’ 

deul – ‘hold, pick 

up’ 

natanae – ‘show, 

present’ 

paagha – ‘identify’ 

figure out’  

georondwe – ‘be 

mentioned, brought 

up’ 

balghyeoji – ‘be 

illuminated’ 

bichu – ‘shine’ 

daebiha – ‘prepare, 

be ready’ 

gamchu – ‘reduce’ 

ganjigha – ‘keep’ 

geojuha – ‘live’ 

ibjiha – ‘be 

positioned at’ 

ilheoga – ‘lose 

something, 

someone’ 

5.09 

5.09 

4.95 

4.90 

4.89 

4.64 

4.60 

4.60 

4.60 

4.60 

4.60 

4.60 

4.60 

4.60 

iljoha – ‘play a part, 

4.60 

contribute’ 

129 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d). 

jeungpogsiki – 

‘amplify’ 

nol – ‘hang out’ 

pyoryuha – ‘drift, 

float’ 

siinha – ‘admit, 

acknowledge’ 

beoseona – ‘get out, 

get free’ 

dayanghaeji – 

‘become diverse’  

simhaeji – ‘become 

severe’ 

saraji – ‘disappear’ 

geod – ‘walk’ 

ganghwaha – 

‘reinforce’ 

barabo – ‘look, 

watch, stare’ 

damul – ‘keep 

quiet’ 

geomtodwe – ‘be 

examined’ 

jeomchi – ‘predict 

future’ 

naebichi – ‘hint at’ 

seongjangha – 

‘grow up’ 

jeonmangha – 

predict 

4.60 

4.60 

4.60 

4.60 

4.60 

4.60 

4.60 

4.49 

4.28 

4.07 

3.94 

3.86 

3.86 

3.86 

3.86 

3.86 

3.85 

130 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d). 

ggeul – ‘pull’ 

georonha – 

‘mention, bring up’ 

jusiha – ‘watch 

carefully’ 

ssah – ‘pile up’ 

gangjodwe – ‘be 

emphasized’ 

hoagboha – 

‘secure’ 

bbomnae – ‘boast, 

show off’ 

gamsiha – 

‘monitor’ 

gangguha – ‘take 

measures to do sth’ 

gongbuha – ‘study’ 

gunrimha – 

‘dominate’ 

pyosiha – ‘express’ 

mat – ‘be in charge 

of something’ 

chireu – ‘pay out’ 

chujeongha – 

‘estimate’ 

jeunggaha – 

‘increase’ 

jis – ‘build, 

construct’ 

mid – ‘believe’  

3.68 

3.48 

3.48 

3.48 

3.26 

3.26 

3.21 

3.21 

3.21 

3.21 

3.21 

3.13 

3.01 

3.00 

3.00 

2.98 

2.98 

2.75 

131 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d). 

jaegiha – ‘raise, 

bring up’ 

balghyeonae – 

‘reveal or disclose’ 

binbalha – ‘occur 

frequently’ 

georaedwe – ‘be 

traded, dealt’ 

giul – ‘lean or tilt’ 

musiha – ‘ignore’ 

neolbhi – ‘make 

wide’ 

maryeonha – 

‘prepare, arrange’ 

ileona – ‘get up’ 

ggob – ‘count (also 

count on fingers)’ 

geodu – ‘reap’ 

gugaha – ‘sing 

praises’ 

jeonragha – ‘fall 

(into ruin)’ 

byeonhwaha – 

‘change’ 

ganjuha – ‘regard, 

consider as’ 

ggal – ‘spread, 

pave’ 

haemyeongha – 

‘clarify’ 

2.69 

2.68 

2.68 

2.68 

2.68 

2.68 

2.68 

2.58 

2.52 

2.47 

2.44 

2.44 

2.44 

2.32 

2.32 

2.32 

2.32 

132 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d). 

hoagdaeha – 

‘expand, enlarge’ 

ibjeungha – ‘prove’ 

insigdwe – ‘be 

acknowledged’ 

naebonae – 

‘remove’ 

hwalyongha – ‘use’ 

unyeongdwe – ‘be 

run, managed’ 

daebyeonha – 

‘represent’ 

ilg – ‘read’ 

seonhoha – ‘prefer’ 

ssodanae – 

‘push/spill out’ 

iru – ‘achieve’ 

bunseogha – 

‘analyze’ 

eongeubdwe – ‘be 

mentioned’  

gureu – ‘stomp 

feet’ 

mangchi – ‘spoil, 

ruin’  

neombo – ‘covet 

(e.g., first place)’ 

soyuha – ‘own’ 

baggu – ‘change’ 

2.32 

2.32 

2.32 

2.32 

2.27 

2.27 

2.26 

2.26 

2.26 

2.26 

2.20 

1.99 

1.93 

1.93 

1.93 

1.93 

1.93 

1.91 

133 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d.) 

jinhaengdwe – 

‘proceed as’ 

chulgandwe – ‘be 

published’ 

chulsidwe – ‘be 

released, launched’ 

ggi – ‘cloud over’ 

gongtongdwe – ‘be 

common’ 

nonha – ‘discuss’ 

majiha – ‘receive, 

greet, welcome 

someone’ 

naseo – ‘take 

action’  

gaebalha – 

‘develop’ 

nae – ‘submit’ 

banyeongha – 

‘reflect’ 

balb – ‘step (on)’ 

deohaega – ‘add’ 

dogryeoha – 

‘encourage’ 

euisimha – ‘doubt 

or be suspicious’  

silgamha – ‘feel, 

sometimes to the 

point of realizing’ 

1.79 

1.63 

1.63 

1.63 

1.63 

1.63 

1.58 

1.55 

1.54 

1.54 

1.54 

1.52 

1.52 

1.52 

1.52 

1.52 

134 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-1 (cont’d).  

jangdamha – 

‘guarantee’ 

teoddeuri – ‘pop, 

break, or burst’ 

ganghwadwe – ‘be 

strengthened’ 

gwasiha – ‘show 

off’ 

naepoha – ‘involve’ 

bijeoji – ‘be made’ 

gajungdwe – ‘be 

aggravated’ 

gamdol – ‘hang’ 

gyeongjaengha – 

‘compete, vie for’ 

haengsaha – 

‘invoke’ 

mangraha – 

‘include or cover 

everything’ 

mojibha – ‘recruit’ 

nanmuha – ‘be rife’ 

1.51 

1.51 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

1.49 

135 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
  
  
 
 
 
 
APPENDIX B: DISTINCTIVE COLLEXEME ANALYSIS II 

Table A-2.  

Distinctive collexemes for the progressive (left) and non-progressive (right) in L1 English L2 

Korean 

Progressive 

Coll.strength 

Non-progressive 

Coll.strength 

saenggagha – ‘think’ 

bo – ‘see’ 

meog – ‘eat’ 

ju – ‘give’ 

bandaeha – ‘oppose’ 

bonae – ‘send’ 

masi – ‘drink’ 

sayongha – ‘use’ 

deud – ‘listen’ 

163.08 

24.43 

8.53 

3.50 

2.86 

2.37 

2.27 

2.27 

1.72 

manhaji – ‘increase’ 

jeunggaha – ‘increase’ 

sal – ‘live’ 

noryeogha – ‘make effort 

baeu – ‘learn’ 

gominha – ‘worry/agonize’ 

jinae – ‘spend/pass time’ 

dani – ‘attend’ 

byeonhwaha – ‘change’ 

geogjeongha – ‘worry’ 

junbiha – ‘prepare’ 

al – ‘know’ 

jeonggongha – ‘major in’ 

neuggi – ‘feel’ 

saenggi – ‘be formed’ 

dwe – ‘become’ 

ilha – ‘work’ 

natana – ‘appear’ 

gongbuha – ‘study’ 

yeonseubha – ‘practice’ 

byeonha – ‘change’ 

gidaeha – ‘expect/anticipate’ 

saraji – ‘disappear’ 

48.37 

42.29 

13.57 

12.47 

11.94 

10.79 

9.49 

8.07 

8.03 

5.86 

5.4 

3.3 

2.92 

2.92 

2.92 

2.84 

2.83 

2.50 

2.23 

1.72 

1.66 

1.66 

1.66 

136 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
APPENDIX C: DISTINCTIVE COLLEXEME ANALYSIS III 

Table A-3.  

Distinctive collexemes for the progressive (left) and non-progressive (right) in L1 Japanese L2 

Korean 

Progressive 

sal – ‘live’ 

Coll.strength 
149.60 

Non-progressive 
saenggagha – 

Coll.strength 
824.74 

138.10 

‘think’ 

ga – ‘go’ 

91.69 

90.75 

moreu – ‘to not 

51.72 

know’ 

75.92 

malha – ‘speak’ 

22.20 

gaji – 

‘have/hold’ 

noryeogha – 

‘make effort’ 

neuleona – 

‘increase’ 

ilha – ‘work’ 

gongbuha – 

‘study’ 

70.02 

69.28 

saenghwalha – 

56.26 

‘live’ 

dani – ‘attend’ 

jinae – ‘spend 

time’ 

40.03 

36.45 

iyagiha – ‘talk’ 

boi – ‘be 

seen/visible’ 

sogaeha – 

‘introduce’ 

o – ‘come’ 

meog – ‘eat’ 

yeonseubha – 

36.22 

bo – ‘see’ 

‘practice’ 

gominha – 

‘worry’ 

dwe – ‘become’ 

gidaeha – 

‘expect’ 

saenggi – 

‘form’ 

32.34 

ju – ‘give’ 

28.46 

24.46 

sa – ‘buy’ 

neuggi – ‘feel’ 

23.47 

sigsaha – ‘eat’ 

137 

19.33 

16.13 

7.97 

4.79 

4.07 

3.43 

3.23 

1.67 

1.60 

1.40 

 
Table A-3 (cont’d). 

sseu – ‘use’ 

balsaengha – 

‘occur’ 

bonae – ‘send’ 

ggeul – ‘pull, 

attract’ 

chaj – ‘find’ 

eungweonha – 

‘cheer’ 

gidari – ‘wait’ 

bad – ‘receive’ 

baljeonha – 

‘develop’ 

moeu – ‘collect’ 

areubaiteuha – 

‘work part-time 

job’ 

ileona – ‘get up’ 

baeu – ‘learn’ 

mid – ‘believe’ 

eod – ‘gain’ 

dallaji –

‘change/become 

different’ 

gareuchi – 

‘teach’ 

saraji – 

‘disappear’ 

mandeul – 

‘make’ 

20.80 

20.38 

19.47 

18.07 

15.33 

15.33 

14.85 

14.60 

12.64 

12.64 

12.47 

12.30 

12.12 

11.12 

9.99 

9.06 

7.41 

7.41 

7.06 

138 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Table A-3 (cont’d). 

baggui – 

‘change’ 

sayongha – ‘use 

hwalyagha – 

‘be active’ 

jjig – ‘take a 

picture’ 

natana – 

‘appear’ 

junbiha – 

‘prepare’ 

nol – ‘play’ 

saraga – ‘make 

a living 

gamsaha – 

‘appreciate’  

baldalha – 

‘develop’ 

haengdongha – 

‘act/behave’ 

jeogeoji – 

‘diminish’ 

silgamha – 

‘realize’ 

deud – ‘listen’ 

pal – ‘sell’ 

sayongdwe – 

‘be used’ 

5.73 

5.14 

4.94 

4.94 

4.44 

3.58 

3.22 

3.22 

2.91 

2.64 

2.64 

2.64 

2.64 

1.95 

1.43 

1.43 

139