Research

This page summarizes my current and past research.

Sentence comprehension and production

My most recent work focuses on cross-linguistic similarities and differences during sentence production and comprehension by Korean and English speakers. I also have investigated processing of aspectual coercion in Korean and English. I'm currently examining the relation between the human behavior in processing aspectual coercion and language models. (expand for full details)

The current understanding in the sentence comprehension literature is that it involves prediction, but we do not yet know how far ahead that prediction spans. In this ongoing project, I probe the timing of verb activation during sentence comprehension in Korean and English. In the context where the verb is predictable from prior context, is it preactivated only before it appears or even before than? I plan to apply the RSA to EEG data collected while participants read scenarios in their native language. I also plan to compare these human data with language models, such as GPT, ACT-R, and Left parsing models.


In a sentence production experiment combined with eye-tracking, I report that both Korean and English speakers rely on the subject for sentence planning. They also looked at the transitive verb image more than the unergative verb, which indicates that they attended to argument structure from the early stage of planning. Meanwhile, differences in the gazes to the verb and the object images by the two groups suggest that the relative weights of object and verb information differ: English speakers mostly rely on the verb while Korean speakers use both verb and object information to encode the argument structure.


Word/morphology processing


I investigate how single word recognition is done both within languages and between languages. Of particular interest is whether grammatical features in different languages have any shared representations. Most recently, I applied machine learning techniques to EEG data (aka neural decoding) to probe lexical and grammatical representations of two languages in English-Korean bilinguals. (expand for full details)

An LDA classifier was trained and tested on EEG data recorded while English-Korean bilinguals read four different nouns and verbs in each language. Above-chance accuracies were obtained in decoding individual lexical item in both languages; the time window for decoding (onset - until approx. 400 ms) suggests that the classification reflects processing of low-level visual properties (~100 ms; Hauk et al., 2006), lexicality (150-200 ms; Pulvermüller et al., 1995), and semantic properties (300-600 ms; Kutas and Hillyard, 1980). Time windows for grammatical tense decoding were different for the two languages, which may reflect different processing mechanisms deployed for inflectional suffixes (English) versus stem conjugation (Korean). Cross-languages decoding, on the other hand, did not yield above-chance accuracies for both lexical items and grammatical features.


Priming effects for morphological roots support theories that grant them a unique representational status in visual word recognition. However, effects for other morphemes, including inflectional affixes, have been inconsistent. In a large-N experiment in English and a Bayesian meta-analysis, I quantify effect sizes of morphological priming of prefix, suffix, and root. The results show that roots and prefixes, but not suffixes, yield morphological priming effects that are statistically larger than form-based priming effects.

 

In this project, I investigate cross-languages priming of grammatical tense from Korean to English, by embedding Korean prime words to English sentences. The results show that lexical items show cross-linguistic masked priming for semantic repetition, especially when primes are in L1. However, grammatical case features do not demonstrate cross-linguistic masked priming effects.

 

Phonological perception 

In this series of research, I investigate phonological perception of Korean consonants & vowels by L1 and L2 Korean speakers with ERPs. In an ongoing project, I investigate how the speaker identity and language experience shape phonological perception. (expand for full details)


In this project, the main question I ask is how English speakers’ prior exposure to Korean accented English and their perceived speaker identity (i.e., whether the speaker is an American or a Korean) modulate their perception of English words with a vowel epenthesis (which is one of key characteristics of Korean-accented English).


Korean aspirated and unaspirated stops are known to be undergoing a VOT merge. Within Korean speakers, older generations (in their 50's) but not younger generations (in their 20's) elicited MMN to the distinction between aspirated (/th/) and unaspirated (/t/) stops. Within Chinese learners of Korean, we report an effect of Korean proficiency in discriminating different Korean stops.

Processing of number agreement by L2 learners [paper]

In my master's thesis, I probe how Korean L2 learners of English process number agreement within NPs using ERPs. I show that English native speakers, but not the L2 learners elicit P600 both when the distance between the determiner (this/they) and the noun (e.g., student/students) is short and long.

Interpretation of Korean modality [paper]

In this paper, I investigate the interaction between the past tense suffix –ess and possibility modal expression –ul-swu-iss in Korean via a semantic acceptability judgment task. The results indicate that this expression is interpreted differently based on the attachment locus of the past tense suffix. When the past tense suffix is attached at the beginning (-ess–ul-swu-iss), it is interpreted as epistemic; when it is attached at the end (-ul-swu-iss-ess), it is interpreted as counterfactual.