Exploring Variability in Compound Tensification in Seoul Korean

Abstract

In Korean noun-noun compounds, the lenis onset consonant in the second noun is often realized as a tense consonant. Although extensive work has been carried out to clarify its causes and relevant phonological processes, this tensification is deemed not entirely predictable. This paper presents a speech production experiment that confirms the existing findings that the variability in tensification is predictable to a certain extent. The experimental results also showed that the relationship between the predictors and the variability is not linear and that tensification mirrors the cognitively determined boundary strength. Native Korean speakers calibrate the boundary strength by incorporating complex information, such as the word's length, segment type, frequency, and plausibility of the compound. While a "tight" boundary led to high tensification probability, it was not affected by speaking-rate variation. Furthermore, the perceived compound's plausibility directly affected the duration of the tensified consonant. Importantly, the findings suggest that speakers' calibration of the boundary strength is fluid and changeable over time and it affects both phonological and phonetic outputs. Finally, variability in data was reduced for the experimental conditions leading to either extremely high or low tensification probability, and there seemed to be lexicalized exceptions to the general trends.

Keywords: Korean; compound; frequency; gemination; tensification.

Figure 1.

Spectrograms with the syllable and constituent word boundaries annotated in /kʰɑpʰ́ɛpɑtɑk/ “café floor.” In (a), the W2 onset consonant is realized as a lenis stop /p/ with voicing throughout the closure. In (b), it is realized as a tense stop /p*/ with a long closure without voicing.

Figure 2.

Averaged compound tensification rate (%) for each participant (participants 1–12) in ascending order.

Figure 3.

Average compound tensification rate (%) for each target compound (x-axis) in ascending order by W2 Length (Monosyllabic vs. Disyllabic) and W1 etymology (Native vs. Loan).

Figure 4.

Boxplots showing the distribution of fitted tensification probability by Etymology, Length and W2 Onset (M = 0.5, SD = 0.08, n = 6,619). The thick line in the box indicates the median, the bottom and the top of the box indicates the 25th and the 75th percentiles, respectively. The whiskers extend to 1.5 times the height of the box, but if there are no data in this range, the whiskers show the maximum or minimum values.