In loanword phonology the effects of L2 (i.e. donor language) prosody or segments on L1 (i.e. a host language) adaptation have been investigated (e.g. Polivanov 1931; 1974; Trubetzkoy 1939; Hyman 1970; Silverman 1992; Paradis & LaCharité 1997; Davis et al. 2012). In the Korean literature, for example, Kenstowicz & Sohn (2001), Lee (2009) and Davis & Lee (2010) among others have examined how English stress or Japanese pitch-accent is borrowed into Kyungsang Korean, and Kang (2003), Ito et al. (2006) and H. Kim (2006; 2007; 2008; 2009; 2014; 2017a; b) have discussed how English, French or Japanese consonants are borrowed into Korean. Some research has extended its focus on how prosodic and segmental factors do interact in L2 adaptation (e.g. Anderson-Hsieh et al. 1992; Munro & Derwing 1999; Cutler 2005; Munro 2008; Caspers & Horłoza 2012). It is in this vein that the present study is concerned with how Japanese H/L tones and segmental voicing contrast affect Korean categorization of Japanese plosives followed by a H or L vowel.

Japanese has a phonemic voicing contrast in plosives, as in (1a) (e.g. /hake/ ‘brush’ vs. /hage/ ‘bald’) (e.g. Shibatani 1990; Shimizu 1996; Tsujimura 1996), and Korean has no voicing contrast but the three-way phonation contrast (i.e. lenis, aspirated and fortis) in plosives which are all voiceless, as in (1b) (e.g. /tal/ ‘moon’, /t^hal/ ‘face mask’, /t’al/ ‘daughter’) (e.g. Kim et al. 2005; 2010; 2018).1 Among the three-way phonation contrast, lenis plosives are often phonetically voiced in intervocalic position (e.g. Silva 1992; Jun 1994; Kim et al. 2018).

Another difference between the two languages is that Japanese is a pitch-accent language, whereas Korean is not except for a few dialects. Japanese has H and L tones of lexical pitch-accent (e.g. McCawley 1968; Haraguchi 1977; 1991; Shibatani 1990; Kubozono 1995; 2011; Tsujimura 1996). For example, the lexical pitch of Tokyo Japanese is a two-way contrast in (a) two-syllable words, a three-way contrast in (b) three-syllable words and a four-way contrast in (c) four-syllable words, when syllables are light, as shown in (2). Words with an accent are marked with ‘'’ after an accented vowel, as in /a'me/. The accented vowels have a H tone with preceding and following vowels having H and L, respectively. Yet, the pitch of the first syllable of a word is low unless the accent is placed on that syllable by virtue of the Initial Lowering Rule (Haraguchi 1977; 1991). The pitch of the first syllable of a Japanese word is also high when the syllable is heavy and has a sonorant in coda position (e.g. /kaisya/ HHH ‘company’; /kantan/ ‘ease’ HHH).

However, Korean has no lexical pitch-accent except for the Kyungsang dialect in South Korea and the Hamkyung dialect in North Korea which are spoken in the sountheastern and northeastern part, respectively, of the Korean peninsula (Ramsey 1978). Like Japanese, Kyungsang Korean which is further divided into two – the North Kyungsang and South Kyungsang dialects – is a pitch-accent language (e.g. K. Chung 1980; G.-R. Kim 1988; Y.-H. Chung 1991; J.-W. Kim 1991; N.-J. Kim 1997; Lee 2009; Lee & Davis 2009, 2010). Thus, accent is lexically assigned to a specific syllable, and pitch patterns are predictable once lexical information such as accent locations or tone register is known in both the South and North Kyungsang dialects. The only difference between the two Kyungsang dialects is that there are no words beginning with LL in South Kyungsang Korean (e.g. Lee & Davis 2009). Unlike Kyungsang Korean, Seoul Korean has no lexical pitch-accent.

On the other hand, it is at a postlexical prosodic level that the H/L tonal distiction plays a role in both Seoul and Kyungsang Korean. According to Jun (1993; 1998; 2005a; b), the Korean three-way laryngeal contrast in plosives is correlated with a boundary tone in the first syllable of the post-lexical prosodic unit called Accentual Phrase (henceforth, AP) in Seoul Korean in intonational phonology (e.g. Pierrehumbert 1980; Beckman & Pierrehumbert 1986; Pierrehumbert & Beckman 1988). That is, words beginning with aspirated or fortis plosives systematically take a H-initial boundary tone and other words take a L-initial boundary tone in AP-initial position (e.g. H tones after the AP-initial aspirated and fortis plosives in /p^halaŋ/ ‘blue’ and /p’alkaŋ/ ‘red’ and a L tone after the AP-initial lenis plosive in /palam/ ‘wind’). Like Seoul Korean which has no lexical pitch-accent, (North and South) Kyungsang Korean has the same consonant-tone correlation in AP-initial position (e.g. G.-R. Kim 1988; Y.-H. Chung 1991; N.-J. Kim 1997; Kenstowicz & Park 2006; Lee 2009), though it has lexical tones as a pitch-accent language like Japanese. For example, Kenstowicz & Park (2006) have found that vowels after AP-initial aspirated and fortis consonants have higher F0 values than those after their lenis counterparts in Kyungsang Korean, as in Seoul Korean. Therefore, regardless of the presence or absence of lexical pitch-accent, Seoul and Kyungsang Koreans are assumed to have boundary tones in AP-initial position.3

Given the differences between Japanese and Korean, we have come to raise two questions. One is how Japanese voiced and voiceless plosives followed by a H or L vowel are categorized into Seoul and Kyungsang Korean dialects in both word-initial and word-medial positions. The other question is how the H/L tonal distinction and segmental voicing contrast in Japanese affects their categorization. In order to explore these issues, a perception experiment was conducted by recruiting Seoul and Kyungsang Korean speakers.

The present study is structured as follows. In Section 2 and 3, the methods and results of our perception experiment are given, respectively; in Section 4 the results of the experiment are discussed; in Section 5 theoretical implications are made; and in Section 6 a brief conclusion is provided.

For the comparison of the Seoul and Kyungsang subjects’ categorization of Japanese plosives followed by a H or L vowel, we conducted a mixed-effect logistic regression model with Dialect (Seoul, Kyungsang), H/L and Dialect by H/L interaction as independent variables and the subjects’ categorization as a dependent variable.

First, as for Seoul subjects’ categorization of Japanese plosives followed by a H or L vowel, we propose that the Korean prosodic unit of AP and laryngeal features do interact, such that the H/L tonal distinction is made, as enhancement, in AP-initial position, and other acoustic properties (i.e. VOT and closure duration) are primarily parsed for cues to Korean laryngeal features.

Given that it is only in word-initial position that the difference in H and L affects the Seoul subjects’ categorization of Japanese plosives, we suggest that the H/L tonal distinction is made in AP-initial position in their adaptation of the Japanese plosives followed by a H or L vowel. Recall that the Japanese test words in (3) were put in the frame sentence /korewa __desu/ ‘This is ___’ and that they were cut from the frame sentence and presented in isolation in the present experiment. The Japanese carrier sentence is phrased as one Intonational Phrase (henceforth, IP) with two APs in a lower prosodic domain, that is, _IP [_AP[korewa] _AP [__desu]] in an autosegmental-metrical model of intonation (e.g. Pierrehumbert 1980; Beckman & Pierrehumbert 1986; Pierrehumbert & Beckman 1988).12 Following the Strict Layer Hypothesis (Selkirk 1986) that the beginning and end of each higher domain is also the beginning and end of lower domains, the word-initial plosives of the Japanese test words presented to Korean subjects are in IP- and hence AP-initial position.13 Thus, our Seoul subjects’ sensitivity to the H vs. L distinction in word-initial Japanese plosives suggests that the Korean prosodic unit of AP plays a role in their categorization of the Japanese plosives, as in the native grammar.

We also propose that other acoustic properties are primarily involved in the Seoul subjects’ categorization of word-initial Japanese plosives with the AP-initial H and L tones as enhancement. As for the other acoustic properties, the examination of Voice Onset Time (VOT) (i.e. the time between the release of a consonant and the onset of voicing in a following vowel) (e.g. Ladefoged 2001) and closure duration of the five-repeated Japanese tokens in (3) revealed that the difference in VOT between word-initial Japanese voiced and voiceless plosives is 31 ms and that in closure duration is 2 ms, as shown in Table 5(a). In the case of word-medial Japanese voiced and voiceless plosives, the difference in closure duration is 28 ms and that in VOT is 9 ms, as in Table 5(b).14

The acoustic properties of VOT and closure duration are articulatorily correlated with glottal opening and linguopalatal or lip contact duration, respectively, in Korean (Kim et al. 2005; 2010; 2018).15 Given that VOT results from glottal opening and that glottal opening peak is significantly higher in aspirated plosives than in non-aspirated plosives (Kim et al. 2018), aspirated plosives are specified as [+spread glottis] (henceforth, [+s.g.]), and lenis and fortis plosives as [–s.g.], as in (4a). The acoustic property of closure duration is accounted for by the feature [±tense], that is, long closure duration in aspirated and fortis stops by [+tense], and short closure duration in lenis ones by [-tense], as in (4b).16

From the data in Table 5 and the primary acoustic cues of VOT and closure duration to the features [±s.g.] and [±tense], respectively, as in (4), we suggest that the difference in VOT between the word-initial Japanese voiceless and voiced plosives is primarily parsed for cues to the Korean laryngeal feature [±s.g.] with the enhancement of the AP-initial H and L tones by virtue of the feature [±tense], as schematized in (5a). That is, in the Seoul subjects’ categorization of word-initial Japanese voiceless plosives as aspirated, long VOT ([+s.g.]) plays a primary role, being parsed for cues to the feature [+s.g.], and the AP-initial H enhances it by virtue of the feature [+tense]. Thus, the Japanese plosives are categorized as aspirated ([+s.g., +tense]), as in (5a i). In the case of their categorization of word-initial Japanese voiced plosives, we suggest that the feature [–tense] is activated as the enhancement of the AP-initial L tone with relatively short VOT primarily parsed for cues to the feature [–s.g.], as in (5a ii). Therefore, word-initial Japanese voiced plosives are categorized as lenis ([–s.g., –tense]).17

As for the Seoul subjects’ categorization of word-medial voiced and voiceless Japanese plosives as lenis and as either aspirated or fortis, respectively, with no tonal effect, we suggest that it is the difference in closure duration that is parsed for cues to the feature [±tense], as in (5b). Long closure duration of word-medial Japanese voiceless plosives is parsed for cues to the Korean laryngeal feature [+tense], and therefore, the voiceless plosives are categorized as either aspirated or fortis ([+tense]), as in (5b i). Short closure duration of word-medial Japanese voiced plosives is parsed for cues to [–tense], as in (5b ii), resulting in their adaptation of the Japanese plosives as lenis.

Second, recall that, like the Seoul subjects, the Kyungsang subjects had better categorization of the word-initial Japanese voiceless plosives as aspirated at each level of syllable length when a tone in a following vowel is H, as shown in Figure 5(b), and word-medial Japanese voiced and voiceless plosives as lenis and as either aspirated or fortis with no tonal effect, respectively, as in Figures 7(b) and 9(b). Therefore, we suggest that the Kyungsang subjects’ categorization of word-initial Japanese voiceless plosives and word-medial Japanese plosives can be accounted for in the same way to the Seoul subjects’, as in (5a i) and (5b), respectively.

Unlike the Seoul subjects, however, the Kyungsang subjects favored H though they categorized word-initial Japanese voiced plosives as lenis like the Seoul subjects. As for the different tonal effects between the Kyungsang and the Seoul subjects’ categorization of word-initial Japanese voiced plosives, we may leave them for further research, because there has been no research of tonal effect on native Korean or loan words of different syllable length in the two dialects in the literature. Therefore, we tentatively suggest that short VOT is primarily parsed for cues to the feature [–s.g.] with the feature [–tense] by default in the Kyungsang subjects’ categorization of word-initial Japanese voiced plosives as lenis.

Before closing this section, we may note that the Seoul and Kyungsang subjects’ categorization of the word-initial Japanese voiceless plosives [k, t] as the aspirated /k^h, t^h/ is of interest in that it is different from the regulations on the Korean adaptation of the word-initial Japanese voiced and voiceless plosives as lenis by the National Institute of the Korean Language (NIKL) (2012). When Japanese words beginning with voiced and voiceless plosives are borrowed and lexicalized in current Korean, the word-initial voicing contrast of the source language is constrained to be neutralized, no matter whether the word-initial plosives are followed by a H or L vowel, as shown in the lexical representations of some Korean adapted forms with word-initial plosives in bold in (6).

The discrepancy between the present results and the NIKL regulations indicates that the regulations, that is, spelling conventions made by NIKL, do not affect our subjects’ categorization of Japanese words at all.18 Recall that we selected real Japanese words which are not familiar at all to Koreans in order to avoid any spelling convention of Japanese loans frequently used in Korea. To be specific, we assume that our Seoul subjects were faithful to the phonetic difference in VOT and in AP-initial L vs. H between the word-initial Japanese voiced and voiceless plosives when given Japanese words which are not lexicalized in Korean. Thus, the acoustic cues are parsed to the features [±tense] and [±s.g.] with the former as an AP-initial enhancement, as in (5a), rather than to the NIKL regulations.19 Another example wherein spelling conventions do not affect our subjects’ adaptation of Japanese words comes from their categorization of word-medial Japanese voiceless plosives. According to the NIKL regulations, we are advised to use aspirated plosives for word-medial Japanese voiceless plosives. However, an almost equal distribution of fortis and aspirated plosives was found in both the Seoul and Kyungsang subjects’ responses for the Japanese voiceless plosives [k, t] in intervocalic position as a free alternation, which is shown in Figure 8(b) and (c).

Some theoretical implications can be made in this section.

First, the present study is the first in examining a cross-dialect difference in the magnitude of H/L tonal effects and segmental voicing contrast in Korean categorization of Japanese plosives followed by a H or L vowel, using two-, three- and four-syllable Japanese words. In the literature on AP-initial tones in accordance with the three-way phonation contrast in plosives, there has been no study which explicitly examines any cross-dialect differences in Korean. So far AP-initial tones have been investigated in the Seoul and Chonnam dialects (e.g. Jun 1993) and in Kyungsang Korean (e.g. Kenstowicz & Park 2006). In addition, there has been no research on how H/L tonal effects and segmental voicing contrast do interact in Korean categorization of Japanese plosives folllowed by a H or L vowel. In this aspect, the present study could be considered as the first attempt to examine a cross-dialect difference in the interaction of Japanese H/L tonal effects and segmental voicing contrast between Seoul and Kyungsang Korean.

Second, the H and L tonal distinction in AP-initial position in the Seoul subjects’ categorization of Japanese plosives indicates that it is not specified in the lexical representation of Korean consonants even among young Seoul Koreans. Recently, Silva (2006) has proposed that lenis plosives are specified as [+s.g.] like aspirated plosives and that H and L are underlyingly specified for aspirated and lenis plosives, respectively, in contemporary Seoul Korean. This is because his three young Korean subjects, who were born after 1982, do not have any difference in VOT between word-initial lenis and aspirated plosives. However, it is noteworthy that among the subjects (80 male and 80 female) in the present study, one hundred forty-seven of them were born between 1985 and 1992 with nine between 1982–1984 and five in 1981. If the tonal distinction were lexically specified in distinguishing lenis from aspirated plosives with no phonological function of the feature [s.g.], then our young Seoul subjects would have perceived the word-initial Japanese voiceless plosives as either aspirated or fortis, in that fortis consonants are also assumed to have H tones like aspirated consonants. But this is not the case.

In addition, the lexical specification of the H and L distinction cannot account for sound patterns, either. Korean, including not only Seoul Korean but also Kyungsang Korean, changes some word-initial lenis consonants into their fortis counterparts in order to convey speakers’ intensified feelings, as shown in (7a).20 No aspirated counterparts occur. The alternation of lenis consonants with fortis counterparts does also occur to the exclusion of aspirated ones in Kyungsang Korean, as in Seoul Korean. For example, as shown in (7b), North Kyungsang Korean also replaces lenis consonants with fortis counterparts in word-initial position, regardless of whether a following vowel is H or L.2122

Moreover, according to Kim et al. (2018), glottal opening peak, airflow peak height, duration of aspiration and VOT of word-medial lenis plosives are significantly reduced, as compared to those in word-initial lenis plosives, in their four subjects of Seoul Korean. In contrast, aspirated and fortis plosives have no statistical difference in them across the contexts. The significant difference in the phonetic properties between word-initial and word-medial lenis plosives is proposed to be due to their laxness ([–tense]), and it is due to the tenseness ([+tense]) of aspirated and fortis plosives that there is no difference in them.

Third, the activation of the Korean prosodic unit of AP as enhancement and laryngeal features [±tense] and [±s.g.] in (5) supports the view that L2 acoustic outputs are computed by an L1 grammar (e.g. Polivanov 1931; 1974; Trubetzkoy 1939; Hyman 1970; Peperkamp 2005; Yip 2006). So far previous studies on the prosodic unit have focused on native or nonsense word segmentation in Korean (e.g. Jun 1993; 1998; 2005a; b; Jun & Kim 2004; S. Kim 2004). In this aspect, the tonal effect in AP-initial position in the Seoul subjects’ categorization of word-initial Japanese plosives provides a case where the prosodic unit of AP is also psychologically real in loanword adaptation, as in native word segmentation. In addition, recently, based on the Korean adaptation of English, French and Japanese voiced and voiceless consonants, H. Kim (2006; 2007; 2008; 2009; 2014; 2017a) has suggested that L1 grammar such as L1 distinctive features, syllable structure and lexical restrictions plays a role in the adaptation of L2 sounds. The present study further supports the view of an L1 grammar-driven perception of L2 sounds, in that Korean AP-initial boundary tones as well as Korean laryngeal features are employed in the Seoul subjects’ categorization of Japanese plosives followed by a H or L vowel.

The other two linguistically-motivated views on loanword adaptation in the literature – (a) the phonological view and (b) the phonetic approximation view – would be hard to account for the results of the present perception experiment. If loanword adaptation is based on phonological category mappings between the L2 and L1 languages (e.g. Paradis & LaCharité 1997; LaCharité & Paradis 2005; Paradis & Tremblay 2009), the H/L tonal distinction in Japanese test words would be borrowed across the contexts by Kyungsang subjects whose dialect has a lexical pitch-accent like Japanese. Yet, this is not the case, as shown in Figures 7 and 9.

The role of the tonal distinction in word-initial position in the Seoul subjects’ categorization (5a) would not be accounted for either in the phonetic approximation view wherein L1 speakers perceive and produce the native segment which most closely approximates the L2 input in articulatory and/or acoustic properties, when confronted with an L2 sound (e.g. Silverman 1992; Steriade 2001; Kenstowicz 2003; Kang 2003; Peperkamp & Dupoux 2003). In this view, the tonal distinction only in word-initial Japanese plosives by the Seoul subjects would not be expected in that the phonetic difference in H and L after Japanese plosives is made in the Japanese acoustic outputs in both word-initial and word-medial positions.23

In the present study, we have found that our Seoul and Kyungsang subjects mostly categorized word-initial Japanese voiceless plosives as aspirated with the significant effect of H and word-medial voiceless plosives as either aspirated or fortis with no H/L effect. Their categorization of word-medial Japanese voiced plosives as lenis is not significantly affected by the H and L tonal difference, either, regardless of dialect differences. In their categorization of word-initial Japanese voiced plosives as lenis, however, the Seoul subjects favored L, and the Kyungsang subjects H.

Based on these results, we have proposed that the Korean prosodic unit of AP and laryngeal features interact in the Seoul subjects’ categorization of word-initial Japanese plosives, such that the H/L tonal distinction is made in AP-initial position as enhancement with VOT primarily parsed for cues to the feature [±s.g.]. Regarding their categorization of word-medial Japanese plosives with no tonal effect, we have proposed that closure duration is parsed for cues to the other laryngeal feature [±tense]. The same proposal is made for the Kyungsang subjects’ categorization except for the H effect in word-initial Japanese voiced plosives which needs further research.

The effect of L1 (i.e. Korean) AP-initial boundary tones as enhancement in the Seoul subjects’ categorization suggests that the prosodic unit of AP is also psychologically real in loanword adaptation, as in native word segmentation, and that the H/L tonal distinction is not underlyingly specified in aspirated and lenis plosives across the contexts in current Seoul Korean. In addition, not only the activation of the L1 AP-initial boundary tones but also that of L1 laryngeal features [±tense] and [±s.g.] supports the view that L2 acoustic ouputs are computed by L1 grammar in the literature. As the first cross-dialect study on the interaction of H/L tonal effects and segmental voicing contrast in Korean categorization of Japanese plosives followed by a H or L vowel, the present study leaves, in the future, further relevant research such as the tonal effects in Seoul and Kyungsang subjects’ perception of native Korean or loan words in different syllable lengths for a supplement to the present study.

The additional files for this article can be found as follows: