Acoustic Data on Vowel Nasalization Across Prosodic Conditions in L1 Korean and L2 English by Native Korean Speakers

Abstract

This article presents acoustic data on coarticulatory vowel nasalization from the productions of twelve L1 Korean speakers and of fourteen Korean learners of L2 English. The dataset includes eight monosyllabic target words embedded in eight carrier sentences, each repeated four times per speaker. Half of the words contain a nasal coda such as p*am in Korean and bomb in English and the other half a nasal onset such as mat in Korean and mob in English. These were produced under varied prosodic conditions, including three phrase positions and two focus conditions, enabling analysis of prosodic effects on vowel nasalization across languages along with individual speaker variation. The accompanying CSV files provide acoustic measurements such as nasal consonant duration, A1-P0, and normalized A1-P0 at multiple timepoints within the vowel. While theoretical implications have been discussed in two published studies, the full dataset is published here. By making these data publicly available, we aim to promote broad reuse and encourage further research at the intersection of prosody, phonetics, and second language acquisition—ultimately advancing our understanding of how phonetic patterns emerge, transfer, and vary across languages and learners.

Dataset: https://doi.org/10.17605/OSF.IO/6WYHR.

Dataset License: CC BY 4.0

1. Summary

The provided CSV files contain acoustic measurements of both carryover and anticipatory vowel nasalization—twelve speakers producing L1 Korean and fourteen producing L2 English. The corresponding acoustic recordings (WAV) are available at OSF [10.17605/OSF.IO/6WYHR], along with these CSV files. The dataset comprises eight monosyllabic target words embedded in eight carrier sentences per language. Each sentence was repeated four times per speaker. Half of the words feature a nasal coda (CVN, C stands for an oral consonant, V for a vowel, N for a nasal consonant)—p*am, p*an, t*am, t*an in Korean (“*” represents a fortis stop in Korean [1]); bomb, palm, den, ten in English. The other half was a nasal onset (NVC)—mat, mak, nat, nap in Korean; mob, mop, Ned, net in English. Productions were elicited under varied prosodic conditions, including three phrase positions (initial for NVC, final for CVN, and medial) and two focus conditions (Focused vs. Unfocused), allowing analysis of prosodic effects on vowel nasalization across languages and speakers. Vowel nasalization (henceforth, V-nasalization) was measured using A1-P0 values [2,3], extracted at multiple timepoints within the vowel. Specifically, measurements were taken at three relative timepoints (25%, 50%, and 75% of vowel duration) and three fixed intervals (20 ms, 40 ms, and 60 ms) from the nasal consonant, providing a comprehensive dataset spanning different portions of the vowel. The CSV files also include acoustic duration measurements of the nasal consonants (henceforth, N-duration) in either onset (for NVC) or coda (for CVN) position in both L1 Korean and L2 English, which allows for the examination of potential correlations between the degree of vowel nasalization and the duration of the nasal consonant.

These data are valuable for researchers investigating acoustic-phonetic variation in nasal consonant duration and vowel nasalization under different prosodic conditions in L1 Korean and L2 English. They are particularly useful for examining how prosodic structure shapes phonetic realization, as well as for exploring the influence of L1 on L2 production. While the original studies [2,3] report results based on nasal duration and normalized A1-P0 values, they do not address individual differences. This dataset therefore offers additional opportunities to analyze speaker-specific strategies in encoding prosodic information, especially among Korean speakers across both L1 and L2 contexts. It also provides a basis for testing how native language experience contributes to variation in L2 production, using the already-measured acoustic data.

2. Data Description

2.1. CSV Files: Acoustic Nasal Duration Measurements

Two separate files containing nasal duration measurements are provided: one for L1 Korean (L1Kor_N-duration.csv) and one for L2 English (L2Eng_N-duration.csv), both produced by native Korean speakers.

Table 1. Sample from the L2 English CSV file (L2Eng_N-duration.csv) for illustrative purposes. The table shows acoustic N-duration measurements for the target word palm, as produced by an advanced Korean learner of English (speaker AF01).

Speaker	Item	Context	Structure	Boundary	Focus	Rep	Duration
AF01	palm	domain-final	CVN	IP	Focused	r1	84.12
AF01	palm	domain-final	CVN	IP	Focused	r2	66.78
AF01	palm	domain-final	CVN	IP	Focused	r3	58.8
AF01	palm	domain-final	CVN	IP	Focused	r4	64.04
AF01	palm	domain-final	CVN	IP	Unfocused	r1	27.05
AF01	palm	domain-final	CVN	IP	Unfocused	r2	64.8
AF01	palm	domain-final	CVN	IP	Unfocused	r3	53.4
AF01	palm	domain-final	CVN	IP	Unfocused	r4	95.88
AF01	palm	domain-final	CVN	Wd	Focused	r1	88.39
AF01	palm	domain-final	CVN	Wd	Focused	r2	59.69
AF01	palm	domain-final	CVN	Wd	Focused	r4	72.52
AF01	palm	domain-final	CVN	Wd	Unfocused	r1	115.38
AF01	palm	domain-final	CVN	Wd	Unfocused	r2	90.77
AF01	palm	domain-final	CVN	Wd	Unfocused	r3	98.86
AF01	palm	domain-final	CVN	Wd	Unfocused	r4	106.3

presents a sample from the L2Eng_N-duration.csv file, which contains acoustic measurements of nasal consonant (N) durations in L2 English. The data include productions of both oral consonant–vowel–nasal consonant (CVN) and nasal consonant–vowel–oral consonant (NVC) word structures, measured for each individual speaker. Each row in the CSV file corresponds to a single token produced under specific experimental conditions. Descriptions of each column, as exemplified in Table 1, are provided below:

• speaker: A unique speaker code identifying each participant. The first letter indicates English proficiency level—”A” for advanced and “I” for intermediate—based on criteria described in Section 3.1 (Participants). The second letter indicates gender (“F” for female, “M” for male), followed by two digits (01–07) randomly assigned to each speaker. Speakers assigned the same identification number in the L1 Korean and L2 English CSV files correspond to the same individuals. Note: AM07 and IM07 are additional participants included only in the L2 English dataset, as the L1 Korean dataset contains 12 speakers, while the L2 English dataset includes 14 speakers.
• item: The target word produced in each trial. For the L1Kor_N-duration.csv file, items include mat, mak, nat, nap, p*am, p*an, t*am, and t*an. For the L2Eng_N-duration.csv file, items include mob, mop, Ned, net, bomb, palm, den, and ten.
• context: The position in which the nasal consonant (N) occurs. domain-final indicates cases where N is the word-final segment (CVN structure), while domain-initial indicates cases where N is the word-initial segment (NVC structure).
• structure: The syllabic structure of the target word, using “C” for oral consonants, “V” for vowels, and “N” for nasal consonants.
• boundary: The prosodic boundary condition under which the target word was produced. “IP” indicates Intonational Phrase boundaries: for NVC words, this means the target occurred phrase-initially (at the beginning of an IP); for CVN words, it occurred phrase-finally (at the end of an IP). “Wd” refers to phrase-medial (word-level) positions.
• focus: The focus condition applied to the target word. “Focused” indicates that the target word was produced under contrastive focus, while “Unfocused” indicates that contrastive focus was placed on another word in the utterance.
• rep: The repetition number for a given condition, labeled from r1 to r4, where “r” stands for repetition.
• duration: The acoustic duration (in milliseconds) of the nasal consonant (N) in the target word.

2.2. CSV Files: Acoustic A1-P0 Measurements at Relative and Absolute Time Points

Four separate CSV files are provided for A1-P0 measurements: two for each language (L1 Korean and L2 English), one with measurements taken from relative time points and the other with measurements taken from absolute time points as labeled below:

• L1Kor_V-nasalization_Absolute.csv
• L1Kor_V-nasalization_Relative.csv
• L2Eng_V-nasalization_Absolute.csv
• L2Eng_V-nasalization_Relative.csv

Each file is organized similarly to Table 1, but with more variables. One notable difference is that CVC (consonant–vowel–consonant) words are added in the item column in these files. The words were included in order to obtain normalized A1-P0 values by examining the range of nasality from the oral to the nasal contexts within each speaker (For the details, see Section 3.3). For this purpose, a CVN or an NVC words were paired with a corresponding CVC word (e.g., bomb vs. bob; and mob vs. bob), as specified in the pair column. Note that both the relative and absolute timepoint files share identical column headings, except for the timepoint column, which differs to reflect the type of temporal measurement used.

• speaker: A unique speaker code identifying each participant (as described in Section 2.1 for acoustic N-duration measurements).
• item: The target word produced in the trial. For L1 Korean, it includes mat, mak, nat, nap, p*am, p*an, t*am, t*an, along with matching consonant–vowel–consonant (CVC) words such as p*at, p*ak, t*at, t*ap, p*ap. For L2 English: mob, mop, Ned, net, bomb, palm, den, ten, and their matching CVC counterparts—bob, bop, dead, debt, pop, Ted.
• pair: The CVC–CVN or CVC–NVC minimal pair to which the target word belongs.
• context: The position in which the nasal consonant (N) occurs (as described in Section 2.1 for acoustic N-duration measurements).
• structure: The syllabic structure of the target word, using “C” for oral consonants, “V” for vowels, and “N” for nasal consonants.
• boundary: The prosodic boundary condition under which the target word was produced (as described in Section 2.1 for acoustic N-duration measurements).
• focus: The focus condition applied to the target word (as described in Section 2.1 for acoustic N-duration measurements).
• rep: The repetition number for each condition, ranging from r1 to r4.
• timepoint: A timepoint for where in the vowel the A1-P0 measurement was taken. For relative timepoint files, values are taken at 25%, 50%, and 75% of vowel duration. In domain-final contexts, 25% is closer to the coda consonant; in domain-initial contexts, 25% is closer to the onset consonant. For absolute timepoint files, values are taken at 20 ms, 40 ms, and 60 ms into the vowel, with proximity interpreted similarly based on the nasal’s position.
• a1p0: The raw A1-P0 value (in dB) measured at the specified timepoint of the token, adjusted based on formant and bandwidth values [4] See Section 3.3 (Measurements) for further detail.
• proportion: The normalized A1-P0 value (%), calculated as a proportion of each target word’s A1-P0 value relative to the nasality range of its minimal pair (CVC–CVN or CVC–NVC). See Section 3.3 (Measurements) for further detail.

2.3. Individual Data Figures

Figure 1, Figure 2, Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 present individual data from fourteen native Korean speakers—twelve of whom contributed to both the L1 Korean and L2 English datasets, with two additional speakers (AM07 and IM07) contributing only to the L2 English data. The figures display nasal consonant duration and normalized A1-P0 values as a function of focus and boundary conditions, plotted separately for each speaker.

• Figure 1 and Figure 2 show nasal consonant duration (in milliseconds) for each of the 12 speakers, categorized by focus type and boundary type, respectively, in L1 Korean.
• Figure 3 and Figure 4 show nasal consonant duration (in milliseconds) for each of the 14 speakers, categorized by focus type and boundary type, respectively, in L2 English.
• Figure 5 and Figure 6 show normalized A1-P0 values (in percentage) at multiple timepoints (both relative and absolute) for each of the 12 speakers, categorized by focus type and boundary type, respectively, in L1 Korean.
• Figure 7 and Figure 8 show normalized A1-P0 values (in percentage) at multiple timepoints (both relative and absolute) for each of the 14 speakers, categorized by focus type and boundary type, respectively, in L2 English.

3. Methods

3.1. Participants

Fourteen native Korean speakers (7 female, 7 male) participated in this production study. Of these, twelve speakers contributed data for both L1 Korean and L2 English, while two additional speakers participated only in the L2 English experiment. All participants were native speakers of Seoul Korean in their twenties (age range: 21–28 years; mean age: 24.7) and had limited overseas experience (less than two years) at the time of recording.

The participants were divided into two groups based on English proficiency: advanced and intermediate learners. The advanced group had TOEFL speaking scores approximately above the 90th percentile or a Level 8 TOEIC Speaking score (the highest level), while the intermediate group had TOEFL speaking scores around the 40th percentile or a Level 6 TOEIC Speaking score (two levels below the highest). For participants who took part in both the L1 Korean and L2 English experiments, the L2 English data were collected approximately one week later (mean interval: 7.7 days) following their participation in the L1 Korean recording session.

3.2. Speech Materials

Eight monosyllabic words with a nasal stop as either onset (mat, mak, nat, nap for Korean; mob, mop, Ned, net for English) or coda (p*am, p*an, t*am, t*an for Korean; bomb, palm, den, ten for English) were used. Matching consonant–vowel–consonant (CVC) words (in the oral context) were also included to determine the range of nasality from the oral to the nasal context for each speaker. Note that NVC and CVN words were used, rather than NVN words, to examine the directionality of the influence from the nasal consonant separately in the onset versus coda context—i.e., the carryover coarticulatory effect in NVC and the anticipatory coarticulatory effect in CVN.

The target words were embedded in a carrier sentence as an answer to a given question in a semi-spontaneous discourse context, as shown in

Table 2. Example sentences with target word mop for the #NVC context and bomb for the CVN# context in English. Target words are underlined, and focused words are in bold.

Structure	Boundary	Focus	Example Sentences
#NVC (mop)	IP-initial	Focused	A: Did you write “bop fast again”? B: No. “Mop fast again” was what I wrote.
		Unfocused	A: Did you write “bop slowly again”? B: No. “Mop fast again“ was what I wrote.
	Wd-initial	Focused	A: Did you write “say bop fast again”? B: No. I wrote, “say mop fast again.”
		Unfocused	A: Did you write “say mop slowly again”? B: No. I wrote, “say mop fast again.”
CVN# (bomb)	IP-final	Focused	A: Were you supposed to write “bob”? B: No. I was supposed to write “bomb”, wasn’t I?
		Unfocused	A: Were you supposed to write “bomb”? B: No. John was supposed to write “bomb”, wasn’t he?
	Wd-final	Focused	A: Did you write “say bob fast again”? B: No. I wrote, “say bomb fast again.”
		Unfocused	A: Did you write “say bomb slowly again”? B: No. I wrote, “say bomb fast again.”

. The mini discourse was designed in such a way to induce different focus types and prosodic boundaries. Table 2 shows example carrier sentences in English, where “B” in each dialogue is the carrier sentence which contains a target word mop for the #NVC context and bomb for the #CVN context.

For instance, in the #NVC context with mop, the IP-initial focused condition is elicited through a question like, “Did you write ‘bop fast again’?” prompting the response, “No. ‘Mop fast again’ was what I wrote.” Here, mop is the focused word, in contrast with bop, and occurs at the beginning of the IP. Conversely, the Wd-initial (=phrase-medial) unfocused condition might involve a question such as, “Did you write ‘say mop slowly again’?” with the response, “No. I wrote, ‘say mop fast again.’” In this case, the contrast is between slowly and fast, so mop is not the focus and occurs phrase-medially (marked as “Wd-initial”). Similarly, in the CVN# context with bomb, the IP-final focused condition could be prompted by, “Were you supposed to write ‘bob’?” leading to the response, “No. I was supposed to write bomb, wasn’t I?” Here, bomb is the focused word, contrasting with bob, and occurs at the end of the IP.

The Korean carrier sentences were also constructed in a similar way, creating distinct prosodic conditions for the target word production, as shown in

Table 3. Example sentences with target word mat for the #NVC context and p*am for the CVN# context in Korean. Target words are underlined, and focused words are in bold.

Structure	Boundary	Focus	Example Sentences
#NVC (mat)	IP-initial	Focused	A: [ipʌn tanʌnɨn p*atsatʃʲin twienonni]? “This time, do I place the word (card) to the right of the picture of p*at?” B: [ani]. IP [matsatʃʲin twi]. IP [twɛssʌ]? “No. To the right of the picture of mat. Got it?”
		Unfocused	A: [ipʌn tanʌnɨn matsatʃʲin aphenonni]? “This time, do I place the word (card) to the left of the picture of mat?” B: [ani]. IP [matsatʃʲin twi]. IP [twɛssʌ]? “No. To the right to the picture of mat. Got it?”
	Wd-initial	Focused	A: [ipʌn tanʌnɨn ap*a p*atsatʃʲin twienonni]? “This time, do I place the word (card) to the right of dad’s picture of p*at?” B: [ani]. IP [a*pa matsatʃʲin twi]. IP [twɛssʌ]? “No. To the right of dad’s picture of mat. Got it?”
		Unfocused	A: [ipʌn tanʌnɨn ap*a mat.satʃʲin aphenonni]? “This time, do I place the word (card) to the left of dad’s picture of mat?” B: [ani]. IP [a*pa matsatʃʲin twi]. IP [twɛssʌ]? “No. To the right of dad’s picture of mat. Got it?”
CVN# (p*am)	IP-final	Focused	A: [ipʌn tanʌnɨn khɨnap*a p*apini]? “This time, is the word (card) big uncle’s p*ap?” B: [ani] IP [khɨnap*a p*am] IP [twɛssʌ]? “No. It’s big uncle’s p*am. Got it?”
		Unfocused	A: [ipʌn tanʌnɨn tʃakɨnap*a p*amini]? “This time, is the word (card) little uncle’s p*am?” B: [ani] IP [khɨnap*a p*am] IP [twɛssʌ]? “No. It’s big uncle’s p*am. Got it?”
	Wd-final	Focused	A: [ipʌn tanʌnɨn khɨnap*a p*ap twienonni]? “This time, do I place the word (card) to the right of big uncle’s p*ap?” B: [ani] IP [khɨnap*a p*am twie] IP [twɛssʌ]? “No. To the right of big uncle’s p*am. Got it?”
		Unfocused	A: [ipʌn tanʌɨn tʃakɨnap*a p*am twienonni]? “This time, do I place the word (card) to the right of little uncle’s p*am?” B: [ani] IP [khɨnap*a p*am twie] IP [twɛssʌ]? “No. To the right of big uncle’s p*am. Got it?”

. As such, this experimental design allows for the systematic examination of how prosodic focus and boundary positions influence the acoustic realization of nasal consonants and vowel nasalization in both L1 Korean and L2 English.

During the experiment, the mini-discourse sentences were presented on a computer screen. A prerecorded prompt sentence of Speaker A (recorded by two native speakers of American English, and by two native speakers of Korean, respectively) was played through a loudspeaker, and the participant answered the question in the role of Speaker B. Acoustic data were collected using a Tascam HC-P2 digital recorder (TASCAM, Santa Fe Springs, CA, USA) and a SHURE KSN44 condenser microphone (SHURE, Wheeling, IL, USA) at a sampling rate of 44 kHz in a sound-attenuated booth.

3.3. Measurements

The acoustic duration of the nasal consonant (N-duration) in CVN# and #NVC was measured from the onset to the offset of the nasal murmur displayed on the spectrogram in Praat [5]. As N-duration includes only the nasal murmur portion, a longer N-duration would indicate enhanced nasality.

The vowel was marked from the end of the onset consonant (i.e., the offset of nasal murmur in NVC and the offset of obstruent aspiration after the burst in CVN) to the onset of the coda consonant (i.e., the offset of vowel formants especially of F2 for NVC and the onset of the nasal murmur for CVN, which is also reflected in substantially weakened formants).

As an acoustic measure of vowel nasalization (V-nasalization), we employed A1-P0, a well-established acoustic parameter used in prior studies (see [4,6,7]). In this measure, A1 refers to the amplitude of the first formant (F1), which reflects the energy concentration around the vowel’s primary resonance, while P0 corresponds to the amplitude of the nasal spectral peak typically located in the 250–450 Hz range, a frequency band where nasal energy is concentrated due to the coupling of the nasal and oral cavities. A larger nasal peak (i.e., lower A1-P0) indicates greater acoustic nasalization.

To capture the temporal dynamics of nasalization across the vowel, A1-P0 values were measured at multiple timepoints within the vowel segment. Specifically, we measured A1-P0 at three relative time points—25%, 50%, and 75% of the vowel duration—allowing us to examine how nasalization evolves over the course of the vowel. In addition, we measured the values at three absolute time points (20 ms, 40 ms, and 60 ms), aligned with the edge of the nasal source: in the #NVC context, time was calculated from the offset of nasal energy from the onset nasal consonant into the vowel; in the CVN# context, time was calculated from the onset of nasal energy from the coda nasal consonant backwards into the vowel. This dual approach enabled a more fine-grained analysis of nasalization patterns around the vowel in both carryover and anticipatory coarticulatory contexts.

All A1-P0 values were extracted using a Praat script [8]. Following the normalization method proposed in [9,10], the raw A1-P0 values were normalized individually for each participant to account for speaker-specific acoustic ranges, as defined below:

Normalized A1-P0 (%):

[(Raw A1-P0 − Minimum A1-P0)/(Maximum − Minimum)] × 100

For this normalization, we used the minimal pair structure of the stimuli (i.e., CVN# vs. CVC# and #NVC vs. #CVC): within each minimal pair, the participant’s maximum and minimum A1-P0 values were identified to determine the participant-specific range of nasalization. Each A1-P0 measurement of the target word was then expressed as a proportion (%) within this range, yielding a normalized value that reflects the degree of nasalization on a comparable scale across speakers. It is also important to note that while raw A1-P0 values are inversely related to the degree of nasalization (i.e., higher A1-P0 values indicate less nasalization due to the dominance of the F1 energy), the normalized values are positively related to nasalization (i.e., higher normalized values correspond to more nasalization), making the interpretation of results more intuitive in the context of our analysis.

This study was conducted in accordance with ethical guidelines and was approved by the ethics committee of the Hanyang Institute for Phonetics and Cognitive Sciences of Language.

4. Findings and Potential Implications

This section outlines key findings from the original research articles [2,3], discusses potential implications of the current data, and suggests future research directions based on the existing dataset.

4.1. Findings and Potential Implications in Relation to L1 Korean Production

• In Korean [2], focus-induced prominence causes nasal consonants to lengthen in domain-final (CVN#), enhancing the [nasal] feature, but shows no such lengthening in domain-initial (#NVC) context; meanwhile, vowels under focus resist nasalization to maintain the [oral] feature in both contexts. Further examination of individual speaker differences as illustrated in Figure 1 reveals that while all 12 speakers consistently lengthen nasal consonant duration in the domain-final (CVN#) context under prominence, only 7 out of 12 speakers do so in the domain-initial (#NVC) context. The remaining five either shorten the nasal duration or show no change under prominence.
• In Korean [2], domain-initial nasals show denasalization with no nasal coarticulatory influence on the vowel, while phrase-final nasals unexpectedly shorten, despite increased vowel nasalization. In the current dataset, out of twelve speakers, nine exhibit shorter nasal duration in phrase-final position, while the remaining three show either longer duration or no change (Figure 2). Additionally, greater individual variation is observed in the vowel nasalization pattern at phrase-final position (Figure 6): six (based on relative time points) and seven speakers (based on absolute time points) show increased vowel nasalization phrase-finally, whereas the other six and five speakers do not show such difference.
• While results from the original study suggest that Korean speakers, as a group, exhibit distinct prosodic fine-tuning compared to English, a closer examination of individual differences therefore reveals the possibility that the specific phonetic mechanisms shaped by prosodic structure may vary across speakers.

4.2. Findings and Potential Implications in Relation to L2 English Produced by Korean Learners

• The L2 English study [2] demonstrates that Korean learners systematically modulate coarticulatory vowel nasalization in English as a function of prosodic structure. These learners appear to access and reproduce prosodic-structural phonetic detail in English with some success; however, their productions remain partially influenced by their L1 Korean, resulting in patterns that diverge from both native Korean and native English norms.
• Our individual speaker data presented in the current data descriptor provide a more nuanced picture of this finding. While most L2 English speakers (13 out of 14) successfully reproduced the lengthened phrase-final nasal consonant in the CVN# context, there was considerable individual variation in how this affected the nasalization of the preceding vowel. Specifically, as can be seen in Figure 8, four speakers (based on relative time points) and six speakers (based on absolute time points) exhibited increased phrase-final vowel nasalization, mirroring native speaker patterns reported in [11]. In contrast, several speakers showed no change, and two exhibited the opposite trend. These findings suggest that, although the aggregate data point to a native-like modulation of coarticulatory vowel nasalization among Korean learners, substantial individual variability persists. This underscores the importance of examining learner data not only in terms of group-level tendencies but also at the individual level to fully capture the complexity of L2 phonetic acquisition.

These potential implications represent just a few examples of what can be drawn from the current dataset. A key theoretical point that emerges from a phonetic perspective is that, although coarticulatory processes—such as vowel nasalization—may originate from physiological and biomechanical constraints of the human speech apparatus [12], they can become internalized within a given language, contributing to cross-linguistic variation. Importantly, the current dataset offers a pivotal insight: such internalized patterns, while shared across a linguistic community, may also be further differentiated at the level of individual speakers’ phonetic systems [13]. Moreover, continued investigation into the varied strategies that individual learners employ to encode prosodic information—both in terms of nasal consonant duration and coarticulatory vowel nasalization—may yield deeper insights into the nature of L2 phonetic acquisition and the interaction between universal articulatory tendencies and language-specific phonetic encoding.

4.3. Further Implications

While previous studies [2,3] primarily focused on the effects of focus and prosodic boundary positions on nasalization patterns, the current dataset—beyond the two exemplified cases discussed above—may also serve as a valuable baseline for future research, offering opportunities for extended analyses in areas including, but not limited to, the following:

• Temporal modification of coarticulatory vowel nasalization: The dataset includes time-aligned acoustic signals that support fine-grained analyses of the temporal extent, onset, and alignment of vowel nasalization in relation to adjacent nasal consonants. This temporal dimension makes it possible to track how nasal coarticulation unfolds over time across different prosodic positions—such as phrase-initial/final versus phrase-medial and focused versus unfocused conditions—providing insight into the dynamic nature of nasalization as a function of prosodic structure. Such analyses can further reveal whether the temporal span of nasalization is systematically adjusted in response to prosodic strengthening or boundary effects, and, crucially, how these patterns may vary across individual speakers, whether in L1 Korean or L2 English contexts.
• Individual variability and L2 prosodic transfer: The dataset is well-suited for investigating speaker-specific variation in coarticulatory patterns, particularly among L2 speakers. It enables analyses of how Korean learners of English differ in their use of coarticulatory vowel nasalization and how their L1 prosodic structure shapes L2 production. Crucially, the dataset includes both native (L1 Korean) and non-native (L2 English) speech from the same individuals, allowing for within-speaker cross-linguistic comparisons. This design offers unique opportunities to examine the influence of L1 phonetic and prosodic systems on L2 speech, especially for non-contrastive features like coarticulatory nasalization—patterns that are not explicitly taught in L2 instruction but may nonetheless be transferred from the speaker’s native language. The inclusion of both Advanced and Intermediate learner groups further facilitates targeted analyses of how L2 proficiency modulates sensitivity to prosodic structure and coarticulatory behavior. Overall, the dataset provides a robust empirical foundation for understanding how language-specific phonetic patterns are internalized, transferred, and potentially reshaped across different levels of linguistic experience.
• Sociophonetic variation (i.e., gender-related differences): The dataset also provides an opportunity to explore sociophonetic dimensions of nasalization, such as gender-based differences in nasalization strategies and their interaction with prosodic structure. For instance, male and female speakers may differ in the extent, timing, or consistency of coarticulatory vowel nasalization, particularly under prosodic prominence or boundary-related modulation. These questions, which have not been addressed in previous publications, offer a valuable direction for future work on how social and linguistic factors intersect to shape fine-grained phonetic variation within and across speech communities.

5. Conclusions

The dataset presented in this data descriptor offers a collection of acoustic phonetic measurements and speech recordings (available via OSF [10.17605/OSF.IO/6WYHR]), focusing on carryover and anticipatory coarticulatory vowel nasalization and nasal consonant duration in both L1 Korean and L2 English, as produced by native Korean speakers. The materials were elicited under systematically varied prosodic conditions—specifically, different boundary positions (Intonational Phrase-initial, medial, and final) and focus conditions (Focused vs. Unfocused)—enabling detailed analysis of how prosodic structure conditions segmental realization. While prior publications based on this dataset [2,3] have reported group-level findings, the full release presented here greatly expands the scope for future research.

A key strength of the dataset lies in its within-speaker cross-linguistic design, which enables direct comparison of L1 and L2 production patterns. With the full set of measured acoustic values and accompanying speech materials, the dataset offers unique opportunities to further investigate the influence of L1 prosodic and phonetic systems on L2 speech—particularly with respect to non-contrastive features such as coarticulatory vowel nasalization. Furthermore, the inclusion of a balanced number of advanced and intermediate English learners makes the dataset well-suited for exploring how L2 proficiency modulates sensitivity to prosodic structure and coarticulatory behavior.

More broadly, the inclusion of individual speaker data is expected to facilitate further investigations into inter-speaker variability—an increasingly central concern in phonetic and phonological research. With the provided data, researchers can explore how Korean speakers differ in encoding prosodic information through nasalization patterns across both native and non-native languages, and how such variation correlates with proficiency level, gender, or speaker-specific strategies. As such, by making these data publicly available, we aim to promote broad reuse and stimulate further research at the intersection of prosody, phonetics, and second language acquisition—ultimately advancing our understanding of how phonetic patterns emerge, transfer, and vary across languages and individual learners.