Variation in loanword adaptation: A case from Mandarin Chinese

Abstract

Mandarin speakers tend to adapt intervocalic nasals as either an onset of the following syllable (e.g. Bruno → bù.lŭ .n uò), as a nasal geminate (e.g. Daniel → dā n.n í.ěr), or as one of the above forms (e.g. Tiffany → dì.fú. n í or dì.fē n.n í). Huang and Lin (2013, 2016) identified two factors that may induce the nasal gemination repair: (1) when stress falls on the pre-nasal vowel and (2) when the pre-nasal vowel is a non-high lax vowel. They hypothesized that Mandarin Chinese speakers insert a nasal coda to perceptually approximate the stronger nasalization and longer syllable duration associated with the stressed syllables, and the shorter vowel duration of a lax vowel because the vowels in closed syllables are shorter in Mandarin. The results from two forced-choice identification experiments and an open-ended transcription task showed that although Mandarin speakers’ choices of different repairs were indeed biased by the different phonetic manipulations, suggesting an effect of perceptual similarity, their decisions were mainly guided by native phonotactics. The overall findings suggest that phonotactic, phonetic, as well as non-linguistic (i.e. frequency) factors interact with each other, resulting in the variable adaptation pattern.

Keywords

English frequency loanword adaptation Mandarin Chinese nasal geminate perceptual similarity phonotactics variation

I Introduction

Emergent patterns that diverge from native patterns are often found in loanword adaptation (e.g. Broselow and Kang, 2013; Kang, 2011). This study is concerned with such a case from Mandarin in which a seemingly unnecessary repair occurs variably in adapting loanwords with intervocalic nasals. Huang and Lin (2013, 2016) report that Mandarin speakers tend to adapt English intervocalic nasals following one of three patterns, with the first being the most common: (1) as an onset of the following syllable (e.g. Bruno → bù.lŭ. n uò), (2) as a nasal geminate (e.g. Daniel → dā n.n í.ěr), or (3) as variation of the first two (e.g. Tiffany → dì.fú. n í or dì.fē n.n í). Since intervocalic nasals are licit in Mandarin, the adaptations with nasal geminates do not seem to stem from the native grammar and appear to be redundant. In Huang and Lin’s corpus studies, they observed that the nasal gemination rate correlated with vowel type and stress location: nasal gemination occurs more often when the pre-nasal vowel is non-high, lax and stressed. They thus resort to a perceptual account and attribute the differential gemination rates to the degree of nasalization and the length of the pre-nasal vowel: non-high pre-nasal vowels are generally more nasalized than high ones, and the shorter vowel duration in VN syllables in Mandarin Chinese is more similar to that of English lax vowels. Furthermore, when pre-nasal vowels are stressed, the total syllable duration is longer. Huang and Lin propose that Mandarin speakers insert a nasal coda to maximize the perceptual similarity to the longer syllable duration in stressed pre-nasal vowels.

Though this perceptual account does seem to solve one piece of the puzzle, it also generates conflicting predictions that cannot be accounted for by corpus data alone. For example, the degree of nasalization is more extensive for tense vowels than for lax vowels, but lax vowels are generally shorter than tense vowels (e.g. Beddor, 2007; Flege et al., 1997; Hillenbrand et al., 1995). Therefore, higher gemination rates would be expected for tense vowels based on their more extensive nasalization, but at the same time, higher gemination rates would also be expected for lax vowels based on their shorter duration. Furthermore, though many previous studies have shown that stressed syllables have longer vowel durations than unstressed syllables in English (Crystal and House, 1988; Klatt, 1976; Van Summers, 1987), the claim that a nasal coda is inserted to match the perceptually longer duration of a stressed syllable does not hold, as CVN syllables are not significantly longer than CV syllables in Mandarin (Wu and Kenstowicz, 2015). Crucially, despite the correlation with stress position and vowel height, the nasal gemination rate in Huang and Lin’s corpus is very low: only 119 out of 2,400 English VNV sequences were adapted with nasal geminates. The majority still follows the singleton N pattern, which suggests that phonetic factors alone could not tell the whole story. One possible factor, not considered in Huang and Lin’s studies, is the phonotactic restriction against the co-occurrence of nasal codas and off-glides (Lin, 2007: 107). This could explain the low nasal gemination rate (i.e. fewer instances of inserting a nasal coda) for syllables with tense vowels. If these vowels were perceived as diphthongs, inserting a nasal coda would be phonotactically illegal.

This study aims to draw on experimental evidence concerning the individual effects of phonetic factors, including nasalization and syllable duration (Experiment 1) and vowel duration (Experiment 2), on the adaptation of English intervocalic nasals in Mandarin Chinese in order to understand the relative contribution of each factor. We further explore how those phonetic factors interact with the phonotactic constraint prohibiting the co-occurrence of a nasal coda and an off-glide (Experiment 3). The experiments reported here use manipulated stimuli to examine one factor at a time to avoid results that may be obscured by an interaction of multiple factors. The results of two forced-choice identification experiments show that, although the phonetic factors do play a role in guiding the speakers in adapting loanwords, there is more to the story. We hypothesize that these phonetic factors interact with phonotactics, a premise that is examined in an open-ended transcription experiment. The results show that such an interaction could indeed account for the variable adaptation patterns. The following section provides background on this particular case from Mandarin Chinese and how previous studies have accounted for it.

II Background

Mandarin Chinese is a tone language with four phonemic tones: high-level Tone 1, rising Tone 2, falling-rising Tone 3, and falling Tone 4. The maximum syllable structure is CGVG/N (Lin, 2007), with either an off-glide or nasal coda, but not both. Huang and Lin (2013, 2016) reported that when English intervocalic nasals (VNV) are borrowed into Mandarin Chinese, one of three possible scenarios occurs (1).

(1) English VNV → Mandarin Chinese (based on Huang and Lin, 2013)

a. Nasal germination 'VNV → VN.NV

Diana → dài.ān.nà

Donald → táng.nà.dé

Helena → hè.lián.nà

Sauna → sāng.ná

Thomas → tāng.mă.sī

b. Singleton nasal 'V:NV → V.NV

Bruno → bù.lŭ.nuò

Sony → suǒ.ní

Conan → kē.nán

Eunice → yoú.ní.sī

V'NV → V.NV

Benet → bèi.nèi.tè

Bonita → bō.ní.tă

Carnegie → kă.nèi.jī

Denise → dí.ní.sī

c. Variation VNV → VN.NV ~ V.NV

Tiffany → dì.fēn.ní ~ dì.fú.ní

Eleanor → āi.lián.nuò ~ āi.lì.nuò

Melanie → méi.lán.ní ~ méi.lā.ní

In (1b), the English intervocalic nasal is adapted as a singleton nasal, whereas in (1a), the adaptation form includes both a nasal coda and a nasal onset, termed ‘nasal gemination’ by Huang and Lin. Note that the definition of ‘gemination’ used here deviates from that in the literature as the nasal coda and the nasal onset are not necessarily identical (e.g. [Vn.mV], [Vŋ.nV], or [Vŋ.mV]). This is partially due to the facts that, in Mandarin Chinese, /m/ is not a licit syllable coda and the vowel and coda in a rime must harmonize in backness (i.e. front vowel +/n/, back vowel +/ŋ/) (Duanmu, 2007; Lin, 2007). Since the nature of the nasals themselves is not the focus of this article, we follow Huang and Lin’s terminology and direct the readers to Hsieh et al. (2009) for a relevant discussion.¹

The different adaptation patterns are proposed to be conditioned by multiple factors. Huang and Lin report that nasal gemination is more likely to occur when the pre-nasal vowel is a non-high lax vowel (i.e. [æ, ɔ, ɛ, ʌ]) that bears the primary stress. Among the 2,400 English proper names with intervocalic nasals in Huang and Lin’s corpus, nasal geminates are found in 119 of the corresponding Chinese loanwords. The majority of these cases is when the pre-nasal vowel is non-high, lax and stressed (91.16%). The remaining cases occur when the pre-nasal vowel is high and tense or when the syllable is unstressed (5.9% and 2.52%, respectively).

They attribute this to the fact that when a pre-nasal vowel is non-high, it has stronger nasalization than when the vowel is high (Bell-Berti, 1993; Henderson, 1984; Krakow, 1994; Solé, 1992). In addition, when a pre-nasal vowel is lax, the syllable has a shorter vowel duration and longer nasal consonant duration compared to syllables with a tense vowel (Beddor, 2007). The duration of vowels in Mandarin Chinese is significantly shorter in CVN syllables than in CV syllables (Wu and Kenstowicz, 2015), which suggests that when Mandarin Chinese speakers adapt an English VNV sequence to VN.NV, they might be perceptually approximating the shorter vowel duration by inserting a nasal coda. As for the stress context, when the pre-nasal vowel bears the primary stress, its duration is longer compared to unstressed vowels (Crystal and House, 1988; Klatt, 1976; Van Summers, 1987). Huang and Lin propose that this is to ‘preserve the longer perceived duration from the stressed syllable in English and preserve the syllable weight in SM [Standard Mandarin] by compensating for a short/lax vowel in English stressed syllables’ (2013: 12). Also, stressed pre-nasal vowels are more nasalized than those without stress (Chen, 2000; Krakow, 1994; Vaissière, 1988a). Mandarin Chinese speakers may insert a nasal coda to match the stronger nasalization since Mandarin vowels are nasalized when preceded by a nasal coda (Chen, 2000).

The effect of perceptual similarity is well-attested in other loanword cases that are also not readily explained by the native grammar (e.g. Boersma and Hamann, 2009; Kawahara and Garvey, 2010; Lu and Hwang, 2015; Peperkamp et al., 2008; Shinohara, 2006). For example, even though final stops are legal in Korean, Korean speakers still variably insert a vowel after final stops in English loanwords. Kang (2003) attributes this to the fact that Korean final stops are unreleased, and the released codas in English are perceptually more similar to their onset counterparts in Korean. The ‘unnecessary repair’ enhances the perceptual similarity between the input form and the adapted form.

Although the perceptual account proposed by Huang and Lin explains up to 91% of the corpus data in which words were adapted with an additional nasal, the phonetic features associated with these factors seem to be weighted differently. First, nasal gemination is said to be triggered when the pre-nasal vowel is non-high due to its stronger nasalization, and lax, for its shorter vowel duration and longer nasal duration, yet lax vowels are more weakly nasalized than tense vowels when preceding a nasal consonant (Beddor, 2007).² Second, if the pre-nasal vowel bears the primary stress, the whole syllable is longer, another factor that may trigger nasal coda insertion. However, according to Wu and Kenstowicz’s aforementioned study, only CGVN syllables are significantly longer than CV, CVN, and CGV syllables in Mandarin Chinese. That is, the duration of CV, CVN, and CGV syllables are not significantly different. The question thus remains if the longer syllable duration associated with primary stress can explain the nasal gemination pattern observed in Mandarin Chinese loanwords. Third, the nasal gemination rate in their corpus is very low. Specifically, less than 5% of English VNV sequences were adapted with nasal geminates. The overwhelming majority is still adapted with a singleton nasal. While phonetic factors have been considered for the role they play in the adaptation pattern, they are not able to tell the whole story. One factor not explored in Huang and Lin’s studies is phonotactics. Specifically, there is a phonotactic restriction against the co-occurrence of a nasal coda and an off-glide in Mandarin (Lin, 2007: 107). This constraint may potentially explain the low nasal gemination rate for tense vowels: Mandarin speakers may perceive the tense vowels as diphthongs, in which case the insertion of a nasal coda would be phonotactically forbidden.

The first two experiments in this article were designed with the aim of gaining a better understanding of the relative contribution of these phonetic factors, specifically nasalization and the duration of vowels and syllables. In the third experiment, we examine the aforementioned phonotactic factor that may account for the variations.

III Experiment 1: Nasalization and syllable duration

The first experiment was designed to investigate the relative contribution of vowel nasalization and syllable duration in triggering nasal gemination in the adaptation of English intervocalic nasals by Mandarin Chinese speakers. A forced-choice identification task was employed to examine Mandarin speakers’ preferences for a form with a nasal geminate or singleton upon hearing an English non-word with an intervocalic nasal.

1 Methodology

a Participants

Twenty native Taiwan Mandarin speakers (12 male, 8 female; aged 20–29 years) were recruited at National Chiao Tung University to participate in this experiment. None of the participants had lived abroad for more than six months, and none reported any hearing deficiencies. All participants were compensated monetarily for their time.

b Design and materials

The stimuli were bisyllabic C₁V₁C₂V₂C₃ sequences, all of which were plausible English non-words, produced naturally by a male native speaker of American English. Recordings were made in a sound attenuated booth using an ATR 1200 microphone at a sampling rate of 44,100 Hz. The pre-nasal vowels consisted of one of the four non-high and non-low English vowels /e [ej], ɛ, o [ow], ɔ/, contrasting in tenseness and backness. Note that the tense vowels used in the experiment were not just long but also diphthongized, as indicated in the narrow transcriptions. This point will become important later on. We chose non-high vowels for their stronger nasalization compared to high vowels (Krakow, 1994; Solé, 1992). C₂ provided either a nasal context, /m/ or /n/, or oral context, /b/ or /d/. Due to phonotactic restrictions in Mandarin Chinese phonology, /ŋ/ and its corresponding oral /g/ were not included in the stimuli. The stress was on either the first or second syllable. Different C₁, V₂ and C₃ segments were provided to increase the variability of the stimuli. None of the stimuli were meaningful words in English or Chinese.

The stimuli were manipulated and divided into four conditions: Stressed, Unstressed, Modified, and No nasal. In the Stressed condition, the stress fell on the pre-nasal vowel, namely, the first syllable (e.g. /dénɛk/). In the Unstressed condition, the stress fell on the second syllable (e.g. /benɪ́p/). Between these two conditions, we predicted that more nasal gemination responses would be given to stimuli in the Stressed rather than the Unstressed condition based on the idea that the overall longer duration of stressed syllables triggers nasal gemination. In the Modified condition, pre-nasal vowels were cross-spliced at the zero crossing using Praat (Boersma and Weenink, 2017) with vowels from corresponding oral contexts, creating pre-nasal vowels without any nasal features (e.g. the first syllable of /ʃénat/ from /ʃédat/). Between the Stressed and Modified conditions, we predicted that the gemination rate in the Stressed condition would be higher due to the fact that the pre-nasal vowels in the Modified condition lacked nasalization. Also, between the Unstressed and Stressed/Modified conditions, we predicted that the gemination rate would be lower for the Unstressed condition since the pre-nasal syllables were shorter in duration.

In each condition, two Vowel Types were included as the pre-nasal vowels: tense and lax. Of these two, pre-nasal tense vowels are characterized by more extensive nasalization. If the intrinsic nasalization of different vowels affects the adaptation pattern, we would anticipate that tense vowels would likely trigger nasal gemination more often than lax vowels.³ The No nasal condition was included to measure baseline performance. In this condition, the stress fell on the first syllable with an oral consonant, /d/ or /b/, in the C₂ position, (e.g. /lédʌʃ/). Nasal gemination was not expected to be triggered in this condition. Examples of the stimuli are provided in Table 1 (for the full list of stimuli, see Appendix 1). The predicted patterns are summarized in Table 2.

Table 1.

Some examples of the stimuli.

Condition	Backness	Tense	Lax
Stressed	Front Back	/dénɛk/ /tómʌf/	/dɛ́nos/ /tɔ́mɪk/
Unstressed	Front Back	/benɪ́p/ /gomɪ́ʃ/	/bɛnʌ́ʃ/ /gɔmʌ́t/
Modified	Front Back	/ʃénat/ (/ʃé/ from /ʃédat/) /sómæp/ (/só/ from /sóbæp/)	/ʃɛ́nɛp/ (/ʃɛ́/ from /ʃɛ́dɛp/) /sɔ́matʃ/ (/sɔ́/ from /sɔ́batʃ/)
No nasal	Front Back	/lédʌʃ/ /tʃóbɛl/	/pɛ́dɔk/ /tʃɔ́bɛs/

Table 2.

Predictions of Experiment 1.

Phonetic factors manipulated	Predicted nasal gemination rates
If longer syllable duration associated with stress has an effect,	Stressed, Modified > Unstressed
If the degree of nasalization has an effect,	Stressed > Modified Tense > Lax

In total, there were 64 stimuli in these four conditions, with 2 Vowel Backness (i.e. front and back vowels), 2 Vowel Tenseness (i.e. tense and lax vowels), 4 Condition (i.e. Stressed, Unstressed, Modified, and No nasal), and 4 tokens in each condition (2 backness × 2 tenseness × 4 conditions × 4 tokens = 64). An equal number of filler items were included to mask the aim of the experiment.⁴

c Procedure

In the forced-choice identification experiment, participants sat individually in a sound attenuated booth using a computer that was connected to a keyboard with two keys labeled ‘1’ and ‘2’. The participants were then presented with written instructions in Chinese on the computer screen. They were instructed to choose an adaptation form provided on the screen that they thought was more perceptually similar to the English non-word that they heard using the keyboard. Each English non-word was presented twice through headphones with a 500 ms inter-stimulus interval (ISI). They were given 6,000 ms to respond before the next trial started. For the experimental items, participants could choose between forms with nasal gemination and without (e.g. hearing /dénɛk/ and given the choices of V.NV and VN.NV). For the filler items, the choices included forms with place contrast (e.g. hearing tubic and given the choice of a dental or velar onset) or aspiration contrast (e.g. hearing pratile and given the choice of an aspirated or unaspirated onset). While these choices were all legal Mandarin Chinese syllables, they were not meaningful words. To avoid any semantic influence from Chinese characters, the choices were given in Bopomofo, a transliteration system for Taiwan Mandarin. Crucially, each pair of choices for the experimental items differed only in whether the first syllable was adapted with a nasal coda or not, or, in the case of filler items, with or without aspiration/place changes. The experimental procedure is illustrated in (2).

(2) Experimental procedure

Prior to the experiment, eight practice trials were presented to make sure that the participants understood the task. After the practice trials, the 128 trials (i.e. 64 stimuli and 64 fillers) were repeated twice with the position of choices counterbalanced and presented randomly in two blocks using E-prime software (Schneider et al., 2002). The whole experiment took about 30 minutes.

2 Acoustic analysis of the experimental stimuli

To ensure that the naturally produced stimuli contained the relevant acoustical correlates, two acoustical analyses of the experimental stimuli were conducted.⁵ The first analysis measured the degree of nasalization of the pre-nasal vowels to ensure that the stressed vowels were more nasalized than the unstressed ones and that the tense vowels were more nasalized than the lax vowels. Second, syllable durations were analysed to ensure that the stressed syllables were longer than the unstressed syllables.

a Degree of nasalization

The degree of nasalization can be measured by A1-P0 or A1-P1 (Chen, 1995, 1997; Styler, 2015). A1 represents the amplitude in dB of the first formant. P0 represents the amplitude in dB of the nasal peak at low frequencies, while P1 represents the amplitude in dB of the nasal peak above the first formant. The difference in amplitude between A1 and P0 is typically negatively correlated with nasalization for non-high vowels, whereas the difference in amplitude between A1 and P1 is negatively correlated with non-low vowels (Chen, 2000). We chose to use A1-P0 rather than A1-P1 as it could be applied more consistently across different types of vowels (Styler, 2015).

The measurements of amplitude difference between A1 and P0 were made at three time points: the beginning, mid-point, and end of the pre-nasal vowels. A1-P0 values were obtained by running the Nasality Automeasure script in Praat⁶ and were then adjusted using the method described in Chen (1997). The mean A1-P0 values for all the stimuli are shown Figure 1, visualized using the ggplot2 package (Wickham, 2009) in R (R Core Team, 2017). The x-axis represents time with three points at which the measurements were taken, and the y-axis represents amplitude, marking the difference between A1 and P0. The graphs are paneled by vowel type.

Figure 1.

Nasalization of the pre-nasal vowels measured by A1-P0 in dB.

From the figure, we can see that the A1-P0 values for Modified (square) and No nasal (cross) conditions were higher than those for the Stressed (circle) and Unstressed (triangle) conditions, meaning there was little to no nasalization in the Modified and No nasal conditions since the target vowels lacked nasal features.

We narrowed our focus to the A1-P0 values at the end of the pre-nasal vowels for Stressed and Unstressed conditions since these values are more representative of the nasalization provided by the nasal context. At the end-point, the pre-nasal vowels in the Unstressed condition had higher A1-P0 values than those in the Stressed condition, indicating stronger nasalization in the Stressed condition vowels, replicating previous findings (Krakow, 1994; Vaissière, 1988b). Furthermore, A1-P0 values of tense vowels (left panel) were lower than those of lax vowels (right panel) at this time point, indicating more extensive nasalization for tense vowels than for lax vowels, a tendency that has also been observed in the literature (Beddor, 2007). A two-way independent-measures ANOVA test confirmed this observation. The effects of Condition (F(1,18) = 3.66, p = .07) and Vowel Type (F(1,18) = 3.3, p = .08) were marginally significant without any interaction (F(1,18) = 0.02, p = .9).

We also measured the duration of nasal consonant and how long nasalization was present in the vowel, as these are also indicators of the degree of nasalization: the shorter the duration of the nasal consonant, the more extensively nasalized the adjacent vowel will be (Beddor, 2007). We did not, however, find a significant difference in nasal consonant durations. That is, the nasals following tense vowels (M = 48.13 ms) were not shorter than those following lax vowels (M = 46 ms). When measuring the duration of vowel nasalization, the pre-nasal vowels were divided into 10 time points, as shown in Figure 2. A comparison of lax and tense vowels revealed similar durations of nasalization. For both types of vowels, there was a significant drop of A1-P0 in the last 10% of the vowel. However, the A1-P0 values for pre-nasal tense vowels were consistently lower than those for pre-nasal lax vowels, an indicator of more extensive vowel nasalization. The same was found for the vowels in the Stressed conditions compared with those in the Unstressed condition, again confirming that the nasalization of the former was stronger.⁷

Figure 2.

A1-P0 values tracked over 10 time points of the pre-nasal vowels in Stressed and Unstressed conditions.

b Syllable duration

For the experimental stimuli, the durations of the first syllables were also measured (in milliseconds); their mean values with standard errors are illustrated in Figure 3.

Figure 3.

Duration of the first syllable in the experimental stimuli.

A one-way independent-measures ANOVA test revealed a significant Condition effect. A post-hoc Tukey HSD test showed that the effect was driven by the longer duration of syllables that carried stress in contrast with those that did not (Stressed vs. Unstressed [p < .01], Modified vs. Unstressed [p = .09] and No nasal vs. Unstressed [p < .05]). This result verifies that the pre-nasal syllables with primary stress were longer than those without stress.

Having confirmed that the acoustical correlates of interest to us were present in the experimental stimuli, the next section presents the results of the first identification experiment.

3 Results

Nasal gemination responses were coded as ‘1’, and those without gemination were coded as ‘0’. Thus, a number closer to 1 indicates a higher nasal gemination rate than a number closer to 0. Figure 4 shows the proportion of nasal gemination responses clustered by vowel type with the x-axis representing the four conditions.

Figure 4.

Results of Experiment 1 testing the effect of nasalization and syllable duration.

Apart from the overall low nasal gemination responses for all conditions (all < 30%), we can see two trends. First, lax pre-nasal vowels received more nasal gemination responses than did tense pre-nasal vowels, and more so in the Modified and Stressed conditions than in the Unstressed and No nasal conditions. Second, more nasal gemination responses were given to stressed vowels compared to unstressed vowels (Stressed/Modified vs. Unstressed). To confirm these tendencies, linear mixed-effects logistic regression modeling was conducted in R (R Core Team, 2017) using the lme4 package (Bates et al., 2015). Models were fitted with Nasal Gemination Responses as the dependent variable and Vowel Type and Condition as the fixed variables. Although both front and back vowels were used in the stimuli, Backness was not included as a fixed variable because it did not improve the model fit significantly based on a Likelihood Ratio Test comparing two models, one with Backness as a fixed effect and one without. The random effects included Stimulus and Participant. In the final model, we included only the random intercept since the models with complex random effect structure failed to converge. In this model, Lax vowel and the Stressed condition were used as the reference levels.

The results showed that nasal gemination responses were higher for lax vowels than for tense vowels in the Stressed condition (β = −1.01, SE = 0.36, p < .01), and this effect was significantly more pronounced in the Modified condition, as indicated by the Tense-Modified interaction (β = −1.52, SE = 0.58, p < .01). Furthermore, nasal gemination rates were higher for lax vowels in the Stressed/Modified conditions (p = .93) compared to the Unstressed condition (β = −0.94, SE = 0.36, p < .01). The gemination rates were the lowest in the No nasal condition (β = −2.29, SE = 0.44, p < .001) compared with those in the referenced Stressed condition. These results indicate that, overall, lax vowels triggered the nasal gemination adaptation pattern more often than tense vowels; this effect was strongest in the Stressed condition, followed by the Modified and Unstressed conditions, and was weakest in the No nasal condition. Moreover, there was a more pronounced difference in nasal gemination rates between tense and lax vowels in the Modified condition than in the Stressed condition, indicating an even lower gemination rate for tense vowels in the Modified condition. The fitted model is shown in (3).

(3) Summary of the fixed effects in Experiment 1

Fixed effects:

Estimate Std. Error z value Pr(>|z|)

(Intercept) -1.51588 0.32158 -4.714 2.43e-06 ***

Unstressed -0.94408 0.36203 -2.608 0.00911 **

Modified 0.03109 0.34053 0.091 0.92725

NoNasal -2.29338 0.43729 -5.245 1.57e-07 ***

Tense -1.00990 0.36290 -2.783 0.00539 **

Unstressed:Tense 0.44529 0.54021 0.824 0.40978

Modified:Tense -1.52146 0.58288 -2.610 0.00905 **

NoNasal:Tense -0.77108 0.88466 -0.872 0.38342

4 Discussion

In Experiment 1, Taiwan Mandarin speakers were asked to choose between forms with or without nasal gemination that best represented the English input that they heard. Nasalization and syllable duration were controlled to varying degrees in the four conditions. Different vowel types (tense and lax) were included to examine the effect of the intrinsic nasalization of these pre-nasal vowels. Furthermore, by comparing the conditions with stressed pre-nasal syllables (the Stressed and Modified conditions) and those without stress (the Unstressed condition), we could detect an effect of syllable duration associated with stress.

The fact that the nasal gemination rate was lower overall in the Unstressed condition than in the Stressed condition indicates that the longer duration associated with stressed syllables does have an effect on Mandarin speakers’ choice of different adaptation forms. We also found higher nasal gemination rates in the Stressed and Modified conditions than in the Unstressed conditions, suggesting that the longer syllable durations induced more nasal gemination responses, regardless of vowel nasalization. These findings confirm Huang and Lin’s hypothesis. However, tense pre-nasal vowels, which have an intrinsically higher degree of nasalization than lax pre-nasal vowels, did not reliably trigger nasal gemination across the four conditions, suggesting that the degree of nasalization associated with different vowel types was overridden by the different stress conditions. That is, nasalization cannot account for why there were across-the-board higher gemination rates for lax vowels than for tense vowels. That being said, we did observe a larger difference between tense and lax vowels in the Modified condition (where the pre-nasal vowels did not contain any nasal features) than in the Stressed condition. This alone suggests an effect of nasalization whereby the tense vowels without any nasal features were given fewer nasal gemination responses compared to those with nasal features. The fact that lax vowels in the Modified and Stressed conditions were not affected by nasalization calls for another explanation, which will be tested in the following experiment.

One may wonder if the higher nasal gemination rates for the lax than for the tense vowels arose from syllable frequencies associated with the stimuli used in the experiment. We calculated the type/token frequencies of the four syllable types used in the experiment based on Taiwan Mandarin conversational corpus (Tseng, 2013), as shown in Table 3 (for the full list of stimuli, see Appendix 1). The results of a one-way chi-squared test showed comparable type frequencies across the four syllable types (χ²(3) = 6.93, p = .07). Although we found higher token frequencies for the front vowels with a CVN structure than with a CV structure ([Cjɛn] > [Cjɛ]), a reversed pattern was observed for the back vowels ([Co] > [Coŋ]). The lack of a vowel Backness effect (see Section III.3) suggests that syllable frequencies of the stimuli used in this experiment were unlikely to bias the participants’ responses (Table 3).

Table 3.

Syllable frequencies of the four syllable types used in Experiment 1.

	CVN	CV	CVN	CV
	[Cjɛn]	[Cjɛ]	[Coŋ]	[Co]
Type frequency	60	44	64	72
Token frequency	24,379	11,491	15,388	33,718

So far, our findings provide some positive evidence for the effect of nasalization and syllable duration associated with stressed syllables triggering more nasal gemination responses to words with intervocalic nasals. The stronger nasalization associated with tense pre-nasal vowels, however, did not elicit more nasal gemination responses. Instead, more responses with nasal gemination were given for lax vowels in each of the four conditions. The lax vowels in the Unstressed condition, however, did not elicit as many nasal gemination responses, which may be attributed to the shorter syllable duration of the unstressed vowels. Duration, it would seem, could therefore play an important role in the adaptation pattern.

It should be noted that though tense vowels have stronger nasalization than lax vowels, the duration of lax vowels is generally shorter than that of tense vowels. Wu and Kenstowicz (2015) reported that the vowel duration of a syllable with a nasal coda is shorter than that of an open syllable in Mandarin Chinese. Therefore, Mandarin Chinese speakers likely insert a nasal coda after lax vowels to perceptually match the shorter vowel duration, as proposed by Huang and Lin (2013, 2016). To better understand the effect of vowel duration associated with vowel type, we designed a second experiment in which the effects of vowel duration was investigated.

IV Experiment 2: Vowel duration

The previous experiment yielded positive results supporting the role of nasalization and syllable duration associated with stressed pre-nasal syllables in guiding Mandarin speakers to give more nasal gemination responses to words with intervocalic nasals. However, we found that, in general, lax vowels triggered more nasal gemination responses than tense vowels across the four conditions, suggesting that the intrinsically stronger nasalization associated with tense vowels did not have an effect. The effect of vowel type thus calls for another explanation. In the second experiment, we explored whether the difference in gemination rates could be accounted for by the shorter duration of lax vowels. Using the same design as the first experiment, the second experiment tested for a duration effect by interchanging the durations of tense and lax vowels to examine if the Mandarin speakers were approximating the shorter duration of lax vowels by inserting a nasal coda (see Wu and Kenstowicz, 2015). Crucially, by employing a between-experiment analysis with the previous experiment in which duration was not manipulated, we could single out the effect of vowel duration.

1 Methodology

a Participants

The same twenty native Taiwan Mandarin speakers who participated in Experiment 1 were recruited for the second experiment to enable a between-experiment comparison. The two experiments were conducted approximately three months apart to avoid any learning effects from the previous experiment. All participants were compensated monetarily for their time.

b Design and materials

The same bisyllabic stimuli from Experiment 1 were used to synthesize the stimuli in this experiment. The vowel durations of the naturally produced stimuli are graphed in Figure 5 below:

Figure 5.

Aggregated vowel durations by vowel type of the experimental stimuli.

Apart from the Unstressed condition in which the vowel durations were shorter overall, a mean difference of approximately 20 ms was observed between the durations of the naturally produced tense and lax vowels, with the former being longer than the latter, as has been consistently reported in the literature (e.g. Crystal and House, 1988; Flege et al., 1997; Hillenbrand and Clark, 2000). In order to examine if the Vowel Type effect observed in Experiment 1 resulted from the shorter duration of lax vowels, the vowel durations in this experiment were manipulated by interchanging the vowel durations of tense and lax vowels in each condition. To ensure a noticeable difference in duration between the two vowel types while maintaining the naturalness of the stimuli, we lengthened each lax vowel and shortened each tense vowel in the Stressed, Modified, and No nasal conditions by 50 ms, using the Pitch Synchronous Overlap and Add (PSOLA) technique in Praat (Boersma and Weenink, 2017). Because unstressed vowels are generally shorter, lax and tense vowels were lengthened and shortened, respectively, by 20 ms in the Unstressed condition. The interchanged durations are shown in Figure 6.

Figure 6.

Aggregated duration by vowel type in the duration-interchanged stimuli.

Here, we see that the lax vowels were longer than the tense vowels in all conditions, and to a greater degree when carrying stress. With these manipulations, the effect of duration can be teased apart from that of nasalization. If shorter vowel duration triggers nasal gemination more consistently, we predict that the opposite pattern from what was found in Experiment 1 will be found in Experiment 2: more nasal gemination responses should be given to the tense vowels rather than the lax vowels due to the interchanged vowel durations (prediction (a) in Table 4). Furthermore, if there is an effect of duration in the Modified conditions between the two experiments, it would be purely due to the durational differences, as nasal features were not involved in this condition; see predictions (b) and (c) in Table 4.

Table 4.

Predictions of Experiment 2.

Phonetic factors manipulated	Predicted nasal gemination rates
If vowel duration associated with vowel type has an effect,	a. Tense > Lax in Experiment 2 (with interchanged duration) b. Tense vowels in Modified condition of Experiment 2 > Tense vowels in the same condition of Experiment 1 c. Lax vowels in Modified condition of Experiment 1 > Lax vowels in same condition of Experiment 2

The predicted patterns are summarized in Table 4. Besides the aforementioned durational manipulations, the stimuli were the same as the previous experiment. The same number of experimental stimuli and filler items were used in this experiment.

c Procedure

The same task and procedure as in Experiment 1 were employed. Prior to the experiment, eight practice trials were presented to make sure that the participants understood the task. After the practice trials, the 128 trials (i.e. 64 stimuli and 64 fillers) were repeated twice with the position of choices counterbalanced and presented randomly in two blocks using E-prime software. The whole experiment took about 30 minutes.

2 Results

We again coded the responses with nasal gemination as ‘1’ and those without as ‘0’. Figure 7 shows the proportion of nasal gemination responses clustered by vowel type with the x-axis representing the four conditions. The results are paneled by Experiment for comparative purposes, with Experiment 1 on the left and Experiment 2 on the right.

Figure 7.

Nasal gemination results paneled by Experiment.

We again used linear mixed-effects logistic regression modeling to interpret the results. Models were fitted with Nasal Gemination Responses as the dependent variable and Experiment, Vowel Type and Condition as the fixed variables. The random effects included Stimulus and Participant. The results from the Modified condition are crucial since the effect of nasalization is controlled for, leaving only the effect of vowel duration. Thus, we set Experiment 2, Lax vowel and Modified condition as the reference levels in the final model (4).

(4) Summary of the fixed effects comparing across Experiment 1 and 2

Fixed effects:

Estimate Std. Error z value Pr(>|z|)

(Intercept) -1.93050 0.37650 -5.128 2.94e-07 ***

Exp 10.36850 0.21996 1.675 0.093867.

Tense -1.01804 0.42645 -2.387 0.016976 *

Stressed 0.44272 0.39181 1.130 0.258505

Unstressed -0.13241 0.40222 -0.329 0.742003

No Nasal -1.77513 0.46928 -3.783 0.000155 ***

Exp1:Tense -1.59968 0.45642 -3.505 0.000457 ***

Exp1:Stressed -0.47078 0.30211 -1.558 0.119164

Exp1:Unstressed -0.94404 0.33453 -2.822 0.004773 **

Exp1:No Nasal -0.58462 0.47485 -1.231 0.218262

Tense:Stressed -0.02497 0.58825 -0.042 0.966141

Tense:Unstressed -0.29364 0.61933 -0.474 0.635413

Tense:No Nasal 0.03723 0.75901 0.049 0.960880

Exp1:Tense:Stressed 1.59987 0.56959 2.809 0.004973 **

Exp1:Tense:Unstressed 2.33671 0.62114 3.762 0.000169 ***

Exp1:Tense:No Nasal 0.75990 1.02577 0.741 0.458806

The results showed that lax vowels in the Modified condition triggered fewer nasal gemination responses in Experiment 2 than in Experiment 1 to a marginally significant level (β = 0.37, SE = 0.22, p = .09). Furthermore, although we also found higher gemination rates for lax vowels than for tense in the Modified condition of Experiment 2 (β = −1.02, SE = 0.43, p < .01), this effect was not as strong as the one found in the same condition in Experiment 1 (β = −1.6, SE = 0.46, p < .001). These results alone suggest a duration effect, in that, compared with the shorter lax vowels in Experiment 1, the longer lax vowels in Experiment 2 discouraged the participants from giving nasal gemination responses. Similarly, the shorter tense vowels in Experiment 2 encouraged the participants to give more nasal gemination responses than the longer tense vowels did in Experiment 1. That being said, the Vowel Type effect (i.e. the stronger tendency to insert a nasal after lax vowels than after tense vowels) was still overwhelmingly evident in all conditions. Possible explanations for this are provided in the following section.

3 Discussion

In the first experiment, we found an overall higher nasal gemination rate in response to lax vowels than to tense vowels. In an attempt to explain this effect of vowel type, Experiment 2 examined the effect of duration by interchanging the intrinsic vowel durations of tense and lax vowels. Although the Vowel Type effect was still evident in Experiment 2, it paled in comparison to that found in Experiment 1, suggesting that, as predicted, the longer vowel duration discouraged the Mandarin speakers to give nasal gemination responses while shorter vowel duration encouraged them to do so. That being said, the effect of duration alone cannot fully account for the adaptation pattern. If it could, we should have seen the opposite pattern between tense and lax vowels since the durations were interchanged between the two categories.

Three important observations can be made from the results of the first two experiments: 1) there was an effect of longer syllable duration associated with stress, in that stressed pre-nasal syllables triggered higher nasal gemination rates than unstressed ones; 2) nasal gemination rates for naturally produced (nasalized) tense pre-nasal vowels were higher than for those in the same context but without any nasal features, suggesting an effect of nasalization; and 3) there was an effect of vowel duration, in that shorter vowels triggered higher nasal gemination rates than longer ones, particularly when comparing the results in the Modified condition in Experiments 1 and 2, though the effect was not strong enough to reverse the overall pattern.

However, the two phonetic factors do not seem to account for the overall pattern: with or without the phonetic manipulations, lax pre-nasal vowels still induced higher nasal gemination rates than tense vowels. Several questions require further investigation. First, it is unclear why the nasal gemination rates were generally low (all conditions < 30%). Second, there is no satisfactory explanation for why we found an effect of nasalization for tense vowels but not lax vowels, as indicated by the comparison of the Stressed and Modified conditions where the only difference was the lack of nasalization of the pre-nasal vowels in the latter. And finally, the overall higher gemination rates for lax vowels than for tense vowels could not be accounted for solely by their shorter duration. If it could, we should have seen an opposite pattern between tense and lax vowels in Experiment 2 in which the vowel durations were interchanged.

For the first question, we can consider the overall preference for open syllables which may be reflected in the frequencies of different syllable types (see Table 5). A calculation of both type and token frequencies of all the legal syllable types in Mandarin based on Taiwan Mandarin Conversational Corpus (Tseng, 2013) indicates that CVN syllables (i.e. V[n] and V[ŋ]) are indeed less frequent than open syllables (i.e. V, VG), and more so for token frequency than for type frequency.

Table 5.

Syllable frequency.

Syllables		Type		Percentage		Token frequency		Percentage
V, GV, VG, GVG		227		58%		451,803		74%
VN	V[n]	165	89	42%	23%	155,213	89,635	26%	15%
VN	V[ŋ]	165	76	42%	18%	155,213	65,578	26%	11%

Notes. In a more standard view, the glides in GV and VG syllables are argued to occupy different syllabic positions: pre-vocalic Gs are governed by Onset and are hence treated as consonants, while post-vocalic Gs are governed by Nucleus along with the main vowel and are hence are treated as vocalic (Duanmu, 2007; Lin, 2007).

While this may account for why Mandarin speakers prefer singleton nasals over nasal geminates in adapting English intervocalic nasals, it does not address why we found across-the-board higher nasal gemination rates for lax than for tense vowels regardless of different experimental conditions.

It is possible that non-phonetic cues, such as phonotactic constraints, play a role here, overriding the effects of the phonetic factors. As mentioned earlier, Mandarin Chinese has a maximal syllable shape of (C)(G)V(G) or (C)(G)V(N). In other words, the co-occurrence of post-vocalic glides and nasal codas is forbidden (i.e. *(C)(G)VGN). For example, it is illegal for the diphthong [eɪ] to be followed by a nasal. Based on this, we hypothesize that the Taiwan Mandarin speakers perceive the tense vowels as diphthongs and inserting a nasal coda is illegal in this context, causing the overall lower nasal gemination rate for tense vowels.

To test this phonotactic hypothesis, we designed another experiment to determine if the participants were more likely to perceive tense vowels as diphthongs thus making them more reluctant to insert a nasal coda when the pre-nasal vowels were tense.

V Experiment 3: Phonotactic effect

This experiment was designed to compare Mandarin participants’ perception of tense and lax vowels in order to determine if the tense vowels were more likely to be perceived as diphthongs (e.g. English /e/ as [eɪ], /o/ as [oʊ]). Participants were asked to transcribe the same stimuli that were used in Experiment 1 and 2 as closely to the audio input as possible using Bopomofo. If perceiving tense vowels as diphthongs contributed to the low nasal gemination rate, the tense vowels should be transcribed as diphthongs more often than lax vowels.

1 Methodology

a Participants

The same group of native Taiwan Mandarin speakers that participated in Experiments 1 and 2 was recruited again for this experiment approximately one month after the second experiment. Of the original 20 participants, 19 were able to return for the third experiment (11 male, 8 female; aged 20–29 years). All participants were compensated monetarily for their time.

b Design and materials

The same bisyllabic sequence of C₁V₁C₂V₂C₃ stimuli from the Stressed and Unstressed conditions from Experiment 1 were used in this experiment, in which the participants were asked to transcribe the first syllable of the words they heard using Bopomofo. There was a total of 32 stimuli (2 vowel backness × 2 tenseness × 2 conditions × 4 tokens).

c Procedure

The transcription task was open-ended, giving participants the freedom to use any combination of Bopomofo, even if it was not a legal sequence. Using this experimental paradigm, the adaptation forms were no longer limited to V.NV and VN.NV as in Experiments 1 and 2. Participants sat individually in a sound attenuated booth and were presented with written instructions in Mandarin Chinese on a computer screen. They were told that after hearing an English non-word, they would see a Bopomofo transcription of the word they heard on the screen with a blank where the target syllable should be inserted.⁸ The participants were then instructed to write down their transcription of the target syllable in Bopomofo on a piece of paper. Participants were given as much time as needed to complete the task. Each English non-word was presented twice with an ISI of 800 ms, and after the participants transcribed their answers on the paper, they would press any key to proceed to the next trial. Prior to the experiment, four practice trials were provided to make sure that the participants understood the task. After the practice trial, the 32 trials were presented in a random order, using E-prime software. The whole experiment took about 30 minutes.

2 Results

The goal of this experiment was to investigate if the strong effect of Vowel Type, regardless of phonetic manipulations, found in the previous experiments was due to tense vowels being perceived as diphthongs more often than lax ones were; if perceived as diphthongs, a phonotactic violation in Mandarin Chinese would cause the lower nasal gemination rates for the tense vowels. For the analysis, the responses of the transcriptions were coded as: 1) diphthongs (e.g. the first syllable of /benɪ́p/ transcribed as [pej] or [paj]), 2) monophthongs (e.g. the first syllable of /benɪ́p/ transcribed as [pɛ], [pjɛ] or [pi]), 3) nasal geminate (the first syllable of /benɪ́p/ transcribed as [pjɛn]or [pən]), and 4) none of the above (/benɪ́p/ transcribed as [p] without any vocalic content).⁹ If the onset was transcribed incorrectly (e.g. /benɪ́p/ transcribed as [dej] or [de]), we ignored the onset and coded their answers based on the vocalic content (e.g. [dej] or [de] would be coded as diphthong and monophthong, respectively). Figure 8 shows the frequency of different types of responses (i.e. Diphthong, Monophthong, Nasal (geminate) and Other).

Figure 8.

Frequency of the coded answers.

We can see from the figure that Diphthong answers were prevalent in response to tense vowels, though they were not uncommon in response to lax vowels. This is an indication that the overall low nasal gemination rates (below 30% in all conditions) in both Experiments 1 and 2 can be partially attributed to phonotactic reasons: Mandarin participants tended to perceive these vowels, independent of vowel tenseness, as diphthongs and inserting a nasal coda in this context creates an illegal syllable in Mandarin. The pattern of Nasal answers follows what was found in the previous two experiments and in Huang and Lin’s work (i.e. more nasal responses were given for lax vowels than for tense vowels, and more so for stressed than for unstressed pre-nasal vowels).

To test our phonotactic hypothesis, we focused our analysis on the Diphthong vs. Monophthong answers for tense and lax vowels. Figure 9 shows that tense vowels were more likely to be perceived as diphthongs than lax vowels were. A two-way chi-squared test with Yates’ continuity correction was used to confirm this tendency with the counts of Diphthong vs. Monophthong answers based on Vowel Type. The results showed that the frequencies of diphthong/monophthong answers in tense and lax vowels were not equally represented (χ² = 36.28 p < .001), suggesting that the proportion of diphthong answers for tense vowels was greater than the proportion of diphthong answers for lax vowels. A comparison of the expected and actual counts for this test is shown in Table 6.

Figure 9.

Proportion of answers by Vowel Type.

Table 6.

Predicted vs. actual counts for the answers based on different vowel types.

		Tense	Lax
Diphthong	Expected	127.75	110.25
Diphthong	Actual	163	75
Monophthong	Expected	164.25	151.75
Monophthong	Actual	129	177

3 Summary

An open-ended transcription test was conducted to further examine whether the strong Vowel Type effect was driven by the perception of English tense vowels as diphthongs, making Taiwan Mandarin speakers reluctant to insert a nasal coda after a tense vowel due to phonotactic restrictions in their native grammar (i.e. *(C)(G)VGN). The results showed that more diphthong responses were given to stimuli containing pre-nasal tense vowels than to those containing lax vowels, supporting our phonotactic hypothesis. That is, Taiwan Mandarin speakers were more likely to perceive tense vowels as diphthongs, and that forbade them from inserting a nasal coda due to phonotactic restrictions in their native language. This phonotactic effect may be the reason why weak phonetic effects were found in Experiments 1 and 2: the perception of tense vowels as diphthongs may have overridden the perception of the phonetic features. Furthermore, perceiving English lax vowels as diphthongs by Taiwan Mandarin speakers were also not trivial (around 30%). This, together with the lower frequency of closed syllables in Mandarin, might be why the nasal gemination rates were all less than 30% in Experiments 1 and 2.

With participants instructed to freely use any combination of Bopomofo, even if they were not legal sequences, one may wonder if the results truly reflect the pure effect of vowel type or if they are instead a by-product of the phonotactic constraints of the Bopomofo symbols, which again reflect the phonotactic constraints of Mandarin. To address this, we analysed the responses in terms of legal combinations and found that, in all the responses, 1,100 were legal and 115 were illegal. Though there was a strong preference for legal combinations, participants still responded with illegal combinations. This, however, should not have obscured the overall results since the nasal insertion responses remained low in this experiment (Figure 8); the effect of the co-occurrence restriction against nasal coda and off-glide should therefore be minimal. The finding that tense vowels were generally perceived as diphthongs while lax vowels were perceived as monophthongs thus still holds.

VI Conclusions

The goal of this study was to provide experimental evidence to explain the variations in Mandarin Chinese in adapting English intervocalic nasals. The results of two forced-choice identification experiments, which tested the effects of nasalization, syllable duration and vowel duration, showed that these phonetic factors only have a limited effect on the adaptation pattern. Instead, a strong Vowel Type effect was found (i.e. more nasal gemination responses for lax than for tense vowels) across the two experiments. We conducted an open-ended transcription task to further examine the phonotactic hypothesis and found that the strong effect of vowel type may have been driven by the perception of English tense vowels as diphthongs, making Taiwan Mandarin speakers reluctant to insert a nasal coda after a tense vowel due to phonotactic restrictions in their native grammar.

The results, taken together, suggest an interaction between phonetic, phonotactic, as well as non-linguistic (i.e. frequency) factors, resulting in the variable adaptation patterns: the stronger nasalization, longer syllable duration and shorter vowel duration encourage a nasal gemination responses while the syllable type frequency and diphthongization of the vowel discourage it. Speakers weighed different factors – nasalization, duration and spectral properties of the pre-nasal vowels – in the adaptation process.

The results also suggest that the perception of spectral properties that caused the tense vowels to be perceived as diphthongs is stronger than that of vowel nasalization and duration, resulting in the overwhelming Vowel Type effect. In other words, when Mandarin speakers were provided with conflicting phonetic cues, for example, stronger nasalization and longer syllable duration of a tense vowel in the Stressed condition, the perception of the spectral properties of a tense vowel still took precedence over nasalization and duration. This can be attributed to the fact that, although different syllable shapes and tones vary in duration (Wu and Kenstowicz, 2015) and vowels can be nasalized to different degrees in nasal contexts (Chen, 2000), duration and nasalization are non-phonemic in Mandarin, while the restriction against the co-occurring off-glide and nasal is deeply rooted in the grammar. To further tease apart perceptual effects from the observed phonotactic effect, a logical follow-up would be to test the adaptation pattern of pre-nasal high vowels, which are less likely to be diphthongized.

Supplemental Material

Appendix – Supplemental material for Variation in loanword adaptation: A case from Mandarin Chinese

Supplemental material, Appendix for Variation in loanword adaptation: A case from Mandarin Chinese by Yangyu Chen and Yu-An Lu in Second Language Research

Footnotes

Acknowledgements

We would like to thank Sang-Im Lee-Kim, Ellen Broselow, the attendees of ICPEAL17-CLDC9, and the Second Language Research reviewers and editors for their valuable comments and ideas. We would also like to thank Yen-Hwei Lin and Ho-Hsin Huang for their work that inspired ours. Any remaining errors are ours.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Higher Education Sprout Project of the National Chiao Tung University, Ministry of Education, Taiwan, and Ministry of Science and Technology of Taiwan Grant (MOST106-2410-H-009-031) to Yu-An Lu.

ORCID iD

Yu-An Lu

Supplemental material

Supplemental material for this article is available online.

Notes

References

Barale

(1982) A quantitative analysis of the loss of final consonants in Beijing Mandarin. Unpublished PhD dissertation, University of Pennsylvania, Philadelphia, PA, USA.

Bates

Maechler

Bolker

, et al. (2015) Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67: 1–48.

Beddor

(2007) Nasals and nasalization: The relation between segmental and coarticulatory timing. In: Trouvain

Barry

(eds) 16th International Congress of Phonetic Sciences. Germany: Saarbrücken, pp. 249–54.

Bell-Berti

(1993) Understanding velic motor control: studies of segmental context. Nasals, Nasalization, and the Velum 5: 63–86.

Boersma

Hamann

(2009) Loanword adaptation as first-language phonological perception. In: Calabrese

Wetzels

(eds) Loanword phonology. Philadelphia, PA: John Benjamins, pp. 11–58.

Boersma

Weenink

(2017) Praat: Doing phonetics by computer [computer program]. Available at: praat.org (accessed September 2020).

Broselow

Kang

(2013) Phonology and speech. In: Herschensohn

Young-Scholten

(eds) The Cambridge handbook of second language acquisition. Cambridge, MA: Cambridge University Press, pp. 529–54.

Chen

(1995) Acoustic parameters of nasalized vowels in hearing-impaired and normal-hearing speakers. Journal of Acoustic Society of America 98: 2443–2453.

Chen

(1997) Acoustic correlates of English and French nasalized vowels. The Journal of the Acoustical Society of America 102: 2360–70.

10.

Chen

(2000) Acoustic analysis of simple vowels preceding a nasal in Standard Chinese. Journal of Phonetics 28: 43–67.

11.

Crystal

House

(1988) Segmental durations in connected-speech signals: Syllabic stress. The Journal of the Acoustical Society of America 83: 1574–85.

12.

Duanmu

(2007) The phonology of Standard Chinese. New York: Oxford University Press.

13.

Flege

Bohn

O-S

Jang

(1997) Effects of experience on non-native speakers’ production and perception of English vowels. Journal of Phonetics 25: 437–70.

14.

Henderson

(1984) Velopharyngeal function in oral and nasal vowels: A cross-language study. Unpublished PhD dissertation, University of Connecticut, Mansfield, CT, USA.

15.

Hillenbrand

Getty

Clark

, et al. (1995) Acoustic characteristics of American English vowels. Journal of Acoustic Society of America 97: 3099–3111.

16.

Hillenbrand

Clark

(2000) Some effects of duration on vowel recognition. The Journal of the Acoustical Society of America 108: 3013–22.

17.

Hsieh

F-f

Kenstowicz

Mou

(2009) Mandarin adaptations of coda nasals in English loanwords. In: Calabrese

Wetzels

(eds) Loan phonology. Amsterdam / Philadelphia, PA: John Benjamins, pp. 131–54.

18.

Huang

Lin

(2013) To geminate or not to geminate: The case of nasals in Standard Mandarin Loanwords. Ann Arbor, MI: NACCL-25, University of Michigan.

19.

Huang

Lin

(2016) When unnecessary repairs become necessary: The case of nasal insertion in Standard Mandarin loanwords. In: Albright

Fullwood

(eds) Proceedings of the 2014 Annual Meetings on Phonology. New York: Linguistic Society of America.

20.

Kang

(2003) Perceptual similarity in loanword adaptation: English postvocalic word-final stops in Korean. Phonology 20: 219–73.

21.

Kang

(2011) Loanword phonology. In: Oostendorp

Ewen

Hume

, et al. (eds) Companion to phonology. Oxford: Blackwell, pp. 2258–82.

22.

Kawahara

Garvey

(2010) Testing the p-map hypothesis: Coda devoicing. New Brunswick, NJ: Rutgers University.

23.

Klatt

(1976) Linguistic uses of segmental duration in English: Acoustic and perceptual evidence. The Journal of the Acoustical Society of America 59: 1208–21.

24.

Krakow

(1994) Nonsegmental influences on velum movement patterns: Syllables, sentences, stress, and speaking rate. In: Huffman

Krakow

(eds) Nasals, nasalization, and the velum: Phonetics and phonology V. New York: Academic Press, pp. 87–116.

25.

Lin

Y-H

(2007) The sounds of Chinese. Cambridge: Cambridge University Press.

26.

Y-A

Hwang

(2015) Phoneme substitution and positional asymmetry in second language learning. In: Hsiao

Wee

L-H

(eds) Capturing phonological shades within and across languages. Newcastle upon Tyne: Cambridge Scholars, pp. 414–32.

27.

Peperkamp

Vendelin

Nakamura

(2008) On the perceptual origin of loanword adaptations: Experimental evidence from Japanese. Phonology 25: 129–64.

28.

R Core Team (2017) R: A language and environment for statistical computing [software]. Vienna: R Foundation for Statistical Computing. Available at: http://www.R-project.org/ (accessed September 2020).

29.

Schneider

Eschman

Zuccolotto

(2002) E-Prime: User’s guide. Pittsburgh, PA: Psychology Software.

30.

Shinohara

(2006) Perceptual effects in final cluster reduction patterns. Lingua 116: 1046–78.

31.

Solé

M-J

(1992) Phonetic and phonological processes: The case of nasalization. Language and Speech 35: 29–43.

32.

Styler

(2015) On the acoustical and perceptual features of vowel nasality. Unpublished PhD dissertation, University of Colorado at Boulder, CO, USA.

33.

Tseng

S-C

(2013) Lexical coverage in Taiwan Mandarin conversation. International Journal of Computational Linguistics and Chinese Language Processing 18: 1–18.

34.

Vaissière

(1988a) Prediction of articulatory movement of the velum from phonetic input. Phonetica 45: 122–39.

35.

Vaissière

(1988b) Prediction of velum movement from phonological specifications. Phonetica 45: 122–39.

36.

Van Summers

(1987) Effects of stress and final-consonant voicing on vowel production: Articulatory and acoustic analyses. The Journal of the Acoustical Society of America 82: 847–63.

37.

Wickham

(2009) Positioning. In: Wickham

(ed) ggplot2. New York: Springer, pp. 115–37.

38.

Kenstowicz

(2015) Duration reflexes of syllable structure in Mandarin. Lingua 164: 87–99.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB