Skip to content
Frontiers
Search

Original Research

Frontiers in Language Sciences ·

Volume 5 - 2026 | https://doi.org/10.3389/flang.2026.1806365

Voice onset time in Arabic L1 learners of English and Spanish: evidence for stability, interaction, and phonetic drift in multilingual speech

  • Hesham Aldamen1
  • Mutasim Al-Deaibes2*corresponding author
  • Rami Alsharefeen3
  • Mohamad Almashour1
  1. 1Department of English Language and Literature, University of Jordan, Amman, Jordan
  2. 2Department of English, American University of Sharjah, Sharjah, United Arab Emirates
  3. 3Faculty Affairs, Rabdan Academy, Abu Dhabi, United Arab Emirates
Article metrics0 viewsView details
PDFCiteMetrics

Abstract

This study examines voice onset time (VOT) in Arabic first-language speakers learning English and Spanish as a multilingual test of the speech learning model-revised (SLM-r). The model predicts continuous phonetic learning, interaction among categories in a shared phonetic space, and bidirectional influence across languages. VOT production was examined in Arabic monolinguals, an Arabic–English bilingual group, an Arabic–English–Spanish multilingual group, and monolingual speakers of English and Spanish using a controlled reading task. Mixed-effects models incorporated group, phonetic condition, and continuous measures of proficiency and relative dominance. English aspirated /p/ was robustly differentiated from English /p/ after /s/ across groups, with a 57.21 ms model-estimated contrast. Spanish /p/ remained in the short-lag range and did not differ reliably between the multilingual group (M = 15.10 ms) and Spanish monolinguals (M = 14.17 ms). For voiced stops, Arabic monolinguals showed substantially more negative Arabic /b/ VOT (M = −85.55 ms) than the Arabic-English (M = −44.08 ms) and Arabic–English–Spanish (M = −43.85 ms) groups, indicating reduced prevoicing in the Arabic L1 learner groups. English /b/ and Spanish /b/ also showed higher prevoicing among Arabic L1 learners than among the respective monolingual controls. Relative English–Spanish dominance did not reliably predict voiceless-stop VOT within the multilingual group. The findings support SLM-r claims about lifelong phonetic plasticity and selective cross-language interaction, while also suggesting that Spanish /p/ may reflect facilitative transfer from Arabic short-lag timing as well as stable category separation.

Keywords

1 Introduction

Second-language speech learning is increasingly understood as a process of continuous phonetic adaptation rather than categorical acquisition followed by stabilization. Contemporary work on bilingual and multilingual speech emphasizes that phonetic categories remain responsive to experience across the lifespan and may be reshaped by sustained use of additional languages (Aldamen and Al-Deaibes, 2023a; Chang, 2019; Flege and Bohn, 2021; Morales-Front, 2018; Zen, 2020). This perspective challenges earlier assumptions that first-language sound systems become fixed after childhood and instead highlights the role of ongoing exposure, use, and dominance in shaping speech production outcomes in bilingual and multilingual speakers. Multilingual learners therefore provide a particularly informative testing ground because they must coordinate multiple phonetic systems that differ in their acoustic targets and phonological contrasts.

Voice onset time (VOT), defined as the temporal interval between stop release and the onset of periodic voicing, has long been treated as a central acoustic correlate of laryngeal timing, that is, the temporal coordination between oral release and vocal fold vibration (Cho et al., 2019; Flege, 1991; Kehoe and Kannathasan, 2021; Lisker and Abramson, 1964). Cross-linguistic research has shown that languages differ systematically in how they realize stop voicing contrasts. Some languages, such as English, rely heavily on the contrast between short-lag and long-lag voiceless stops, whereas others, such as Spanish and many Arabic varieties, rely more heavily on the contrast between prevoiced stops and short-lag voiceless stops (Al-Tamimi et al., 2021; Flege and Eefting, 1987; Rosner et al., 2000; Al-Deaibes et al., 2026). Because VOT directly indexes these cross-language differences, it has remained a central measure in research on phonetic category formation, bilingual speech production, and multilingual interaction (Davis, 1995; Goldrick et al., 2021; Lowenstein and Nittrouer, 2008; Macken and Barton, 1980).

Arabic, English, and Spanish provide a particularly useful language triad for testing multilingual phonetic interaction. English typically contrasts long-lag aspirated voiceless stops with relatively short-lag voiced stops in word-initial position. Spanish voiceless stops are generally short-lag and minimally aspirated, whereas voiced stops are widely described as robustly prevoiced, especially in Castilian benchmarks (Flege and Eefting, 1987; Lisker and Abramson, 1964; Rosner et al., 2000). Arabic stop systems show dialectal variation, but many descriptions likewise report short-lag voiceless stops and substantial prevoicing for voiced stops (Al-Tamimi et al., 2021). For Arabic-speaking learners, English therefore introduces a laryngeal timing pattern that is distinct from Arabic, while Spanish partially overlaps with Arabic for voiceless timing but still differs in lexical, phonological, and multilingual representational status. These relationships make the language triad theoretically informative for examining both category separation and cross-language facilitation.

Longitudinal and training research on L2 English and Spanish further shows that learners can move toward target-like VOT over time, although development is typically non-linear and characterized by substantial individual variation (Munro et al., 2024; Nagle, 2019; Olson, 2019, 2021). These studies reinforce the view that VOT is especially useful for examining how cross-language similarity, input distribution, and learning history shape bilingual and multilingual phonetic outcomes.

Prior work provides several relevant benchmarks but leaves important gaps. Early research documented cross-language phonetic interference from Arabic to English stop production (Flege and Port, 1981). Research on Spanish-speaking learners of English and on Spanish stop timing has established the centrality of prevoicing and short-lag voiceless stops in Spanish benchmarks (Flege and Eefting, 1987; Rosner et al., 2000). Studies of bilingual and multilingual VOT production further show that speakers can maintain language-specific stop timing while still exhibiting cross-language interaction that depends on language context, dominance, structural similarity, and category stability (Antoniou et al., 2010, 2011; Geiss et al., 2022; Olson, 2013; Souganidis et al., 2024; Stoehr et al., 2024). Arabic remains comparatively underrepresented in this literature, especially in designs that compare Arabic L1 bilingual and multilingual groups within the same analytic framework (Al-Damen et al., 2025; Hamzah et al., 2020).

The present study addresses this gap by examining VOT production in Arabic monolinguals, an Arabic–English bilingual group, an Arabic–English–Spanish multilingual group, and monolingual speakers of English and Spanish. The study tests SLM-r predictions concerning continuous phonetic learning, shared phonetic space, cross-language interaction, and L2-to-L1 phonetic drift by combining controlled production data with mixed-effects modeling and continuous measures of proficiency and dominance. It also responds directly to the need for clearer Arabic-specific phonetic contextualization in multilingual phonetics, including how Arabic stop timing aligns with and diverges from English and Spanish (Aldamen and Al-Deaibes, 2023a,b; Flege and Bohn, 2021).

The study is guided by the following research questions and accompanying predictions. First, we asked whether Arabic L1 learners differentiate VOT across English aspirated voiceless stops, English unaspirated voiceless stops in s-clusters, and Spanish voiceless stops. We predicted clear maintenance of the English aspiration contrast and short-lag realization of Spanish /p/. Second, we asked whether simultaneous acquisition of English and Spanish leads to cross-language phonetic interaction. We predicted little evidence of systematic over-aspiration in Spanish, while acknowledging that target-like Spanish /p/ could also reflect facilitative transfer from Arabic short-lag timing. Third, we asked whether proficiency and relative dominance in English and Spanish predict gradient VOT variation in the multilingual group. We predicted limited dominance effects for these relatively stable voiceless-stop categories. Fourth, we asked whether there is evidence of L2-to-L1 phonetic drift in Arabic /b/. We predicted less negative Arabic /b/ VOT in the Arabic L1 learner groups than in Arabic monolinguals.

  • Do Arabic L1 learners differentiate voice onset time across English aspirated voiceless stops, English unaspirated voiceless stops in s-clusters, and Spanish voiceless stops?
  • Does simultaneous acquisition of English and Spanish lead to cross-language phonetic interaction, such as over-aspiration in Spanish voiceless stops or reduced aspiration in English voiceless stops?
  • To what extent do proficiency and relative dominance in English and Spanish predict gradient variation in voice onset time among Arabic L1 multilingual learners?
  • Is there evidence of L2-to-L1 phonetic drift in the production of Arabic voiced stops when compared with Arabic monolingual speakers?

2 Literature review

The speech learning model-revised proposes that phonetic categories across a speaker's languages occupy a shared representational space and remain plastic across the lifespan (Flege and Bohn, 2021). Under this view, new phonetic categories are not learned in isolation. Instead, their development depends on perceived similarity to existing categories, the distribution of input, and patterns of language use. Shared space does not require categorical merger, but it does predict that categories can interact, shift, and sometimes drift as multilingual experience accumulates. VOT has become one of the clearest testing grounds for these claims because it provides a directly measurable correlate of laryngeal timing that is comparable across languages (Cho et al., 2019; Flege, 1991).

Research applying SLM-compatible logic to stop production has shown that bilinguals often maintain language-specific VOT patterns while also exhibiting structured cross-language influence. Language context can elicit native-like stop voicing in both languages, particularly when categories are well established and sufficiently separated functionally (Antoniou et al., 2010). At the same time, bilingual and multilingual speakers may show inter-language interference and asymmetrical phonetic code-switching when language control demands increase or when structurally similar categories compete for implementation (Antoniou et al., 2011; Olson, 2013). These findings are especially relevant for multilingual VOT because they show that phonetic separation and interaction can coexist rather than excluding one another.

Arabic-speaking learners contribute an important but still limited body of evidence to this literature. Flege and Port (1981) documented cross-language phonetic interference from Arabic to English, establishing that Arabic stop timing can shape English stop realization. More recent studies of Arabic-speaking learners of English have likewise shown that aspiration can be acquired successfully, although variability remains linked to experience, proficiency, and language use (Al-Damen et al., 2025; Hamzah et al., 2020). Research on Arabic phonetic learning outside the stop system also supports the broader claim that adult learners can develop new phonetic targets while still carrying forward L1-based constraints and category interaction (Aldamen and Al-Deaibes, 2023a,b).

Evidence for L2-to-L1 phonetic drift is also relevant to the present design. Studies of adult bilinguals have shown that sustained use of an additional language can shift native-language phonetic behavior, sometimes even over relatively short periods of exposure (Chang, 2012, 2013, 2019; Kornder and Mennen, 2021; Stoehr et al., 2017). In Arabic-related contexts, Alharbi et al. (2023) reported L2 acquisition and L1 attrition effects in highly proficient late bilinguals, and Kelly et al. (2021) showed that bilingual language mode can influence VOT realization. These findings support the SLM-r expectation that L1 categories remain modifiable rather than fixed.

Research on Spanish stop timing supplies another crucial benchmark. Spanish voiceless stops are widely described as short-lag and minimally aspirated, whereas Spanish voiced stops are typically associated with robust prevoicing and strongly negative VOT values in canonical descriptions (Face and Menke, 2020; Flege and Eefting, 1987; Lisker and Abramson, 1964; Rosner et al., 2000). For Arabic speakers, this creates a potentially facilitative relationship for Spanish /p/ because Arabic voiceless stops are also commonly short-lag. It also creates an important interpretive question for Spanish /b/: if Spanish and Arabic both support substantial lead voicing, then reduced Arabic /b/ prevoicing in multilingual learners cannot be attributed straightforwardly to Spanish exposure alone.

Methodologically, recent VOT research increasingly models proficiency, dominance, and experience as continuous variables rather than categorical labels (Geiss et al., 2022; Souganidis et al., 2024; Takahashi, 2023). This approach is well suited to multilingual phonetics because it allows category interaction to be evaluated alongside individual differences in experience. Even so, few studies have directly compared Arabic–English bilinguals with Arabic–English–Spanish multilinguals using matched materials, monolingual comparison groups, and unified mixed-effects analyses. The present study addresses that gap while also clarifying the limits imposed by a single lexical item per condition and by the absence of an Arabic voiceless bilabial comparison item.

3 Method

3.1 Design and overview

The study employed a mixed experimental design to investigate voice onset time in Arabic, English, and Spanish. Condition, defined by language and stop category, was treated as a within-speaker factor, and Group, defined by language background, was treated as a between-speaker factor. The primary dependent variable was VOT (ms), defined as the temporal interval between stop release and the onset of periodic voicing. For voiced stops, VOT values could be negative when voicing preceded release. In addition, voiced stops were coded for prevoicing presence (0 = absent, 1 = present) and, when present, prevoicing duration (ms).

Six experimental conditions were analyzed, as coded in the dataset: English /p/ in a simple onset (Eng_p_asp), English /p/ after /s/ in s-clusters (Eng_p_after_s), English /b/ (Eng_b), Spanish /p/ (Spa_p), Spanish /b/ (Spa_b), and Arabic /b/ (Arb_b). This design enabled direct comparison of aspirated, short-lag, and prevoiced stop realizations across monolingual, bilingual, and multilingual speakers and supported tests of cross-language interaction, dominance effects, and potential L2-to-L1 phonetic drift.

3.2 Participants

The final sample consisted of 125 adult participants aged 18–25 years, divided evenly across five groups (n = 25 per group): Arabic monolinguals, Arabic–English bilinguals, Arabic–English–Spanish multilinguals, English monolinguals, and Spanish monolinguals. All participants reported normal hearing and no history of speech, language, or neurological disorders. Mean age and standard deviation by group were as follows: Arabic monolinguals (M = 21.56, SD = 2.50), Arabic-English participants (M = 21.32, SD = 2.26), Arabic–English–Spanish participants (M = 20.96, SD = 2.17), English monolinguals (M = 21.92, SD = 2.57), and Spanish monolinguals (M = 21.04, SD = 2.28). Ages ranged from 18 to 25 years across all groups. Gender distribution by participant count was as follows: Arabic monolinguals included 12 females and 11 males; Arabic–English participants included 10 females and 13 males; Arabic–English–Spanish participants included 12 females and 11 males; English monolinguals included 13 females and 11 males; and Spanish monolinguals included 15 females and 10 males.

The Arabic-speaking groups were recruited in the institutional context in which the recordings were collected under University of Jordan ethics approval. The English and Spanish monolingual groups served as age-matched comparison baselines for the target languages. Detailed hometown-level provenance and current-residence metadata were not retained in the anonymized analysis file for all control speakers. For that reason, the monolingual groups are interpreted here as empirical comparison baselines for the present dataset rather than as dialectally exhaustive norms. This point is especially relevant for the Spanish /b/ baseline discussed below.

A formal a priori power analysis was not conducted because the dataset had been completed before the present confirmatory analysis. The design nevertheless includes 25 speakers per group and repeated productions within speaker, which provides a reasonable basis for detecting medium-to-large group differences in participant-level mixed-effects models. Because the materials include a single lexical item per condition, the design was not intended to support fine-grained lexical generalization.

3.3 Language background, proficiency, and age of acquisition

Language proficiency was measured using departmental standardized tests and recorded as percentage scores. English proficiency scores averaged 83.56 (SD = 5.42, range = 75.03–91.96) for the Arabic–English group, 85.31 (SD = 4.99, range = 78.29–93.18) for the Arabic–English–Spanish group, and 95.84 (SD = 2.31, range = 92.29–99.93) for English monolinguals. Spanish proficiency scores averaged 71.97 (SD = 7.96, range = 60.01–84.60) for the Arabic–English–Spanish group and 94.27 (SD = 3.37, range = 90.11–99.85) for Spanish monolinguals.

Age of acquisition (AoA) was reported in years. For English, mean AoA was 8.72 (SD = 1.87) for the Arabic–English group and 9.16 (SD = 2.24) for the Arabic–English–Spanish group. For Spanish, mean AoA in the Arabic–English–Spanish group was 17.04 (SD = 2.17). English monolinguals and Spanish monolinguals were coded as having an AoA of 0 for their respective native language.

For the Arabic–English–Spanish group, English and Spanish proficiency scores were z-standardized within the group to create Eng_z and Spa_z. A continuous language dominance index was then calculated as the difference between the two standardized scores (Balance = Eng_z – Spa_z). The Balance index had a standard deviation of 1.37 and ranged from −2.81 to 2.77, with positive values indicating greater relative dominance in English and negative values indicating greater relative dominance in Spanish.

3.4 Speech materials and procedure

The dataset contained one target word per condition. For English, the target words were pat (Eng_p_asp), spat (Eng_p_after_s), and bat (Eng_b). For Spanish, the target words were pata (Spa_p) and bata (Spa_b). For Arabic, the target word was baat (Arb_b). Each item was produced three times by each participant in the relevant group or groups.

The one-item-per-condition design was selected to maximize cross-language phonetic comparability and to minimize lexical and prosodic variability across the three languages. The trade-off is that inference is restricted to participant-level patterns rather than lexical generalization, a point reflected in the statistical modeling below. Arabic voiceless /p/ was not included because the Arabic baseline comparison targeted native Arabic stop categories represented consistently across participants. Introducing a borrowed or non-native Arabic /p/ item would have added lexical familiarity and category-status confounds. At the same time, the absence of an Arabic voiceless bilabial comparison limits how directly the present data can separate Spanish /p/ category formation from facilitative transfer from Arabic short-lag voiceless timing.

Participants completed a controlled reading task in their assigned language or languages. Each target word was elicited three times in separate trials, and no additional self-paced repetitions were permitted beyond these planned repetitions. This procedure yielded a total of 1,200 tokens prior to exclusion. Token counts by group were 75 tokens for Arabic monolinguals, 225 tokens for English monolinguals, 150 tokens for Spanish monolinguals, 300 tokens for Arabic–English bilinguals, and 450 tokens for Arabic–English–Spanish multilinguals. Recordings were made using a head-mounted microphone to ensure stable mouth-to-microphone distance. Sessions were conducted in a quiet environment suitable for acoustic analysis. Audio files were digitized and stored for offline annotation.

3.5 Acoustic annotation and data cleaning

VOT was measured in milliseconds using waveform and spectrogram displays in Praat, following standard practice in acoustic phonetics (Cho et al., 2019; Flege, 1991; Lisker and Abramson, 1964). Stop release, or burst, was identified at the abrupt release transient visible in the waveform and spectrogram. Voicing onset was identified at the first regular glottal pulses following release. For voiced stops, negative values were retained when periodic voicing began during the closure interval before release. In the present manuscript, prevoicing refers to this closure-period voicing before release, short lag refers to brief positive VOT following release, and long lag refers to the longer postrelease interval typically associated with aspiration.

For /b/ tokens, prevoicing presence was coded as 1 when periodic voicing was visible during the closure interval immediately preceding release and 0 otherwise. When present, prevoicing duration (ms) was measured. Aspiration was identified acoustically as an extended positive VOT interval with postrelease aperiodic noise before regular voicing onset.

Each token was marked with an exclusion flag. Tokens were excluded if they contained misread items, unclear burst release, ambiguous voicing onset, or recording artifacts. Unclear burst release included masked or multiply released tokens in which a unique release point could not be assigned reliably. Ambiguous voicing onset referred to tokens with irregular, interrupted, or visually indistinct onset of periodicity. Recording artifacts included clipping, transient noise, or signal degradation that prevented reliable measurement. Across the dataset, 26 of 1,200 tokens were excluded, yielding an exclusion rate of 2.17%. All analyses were conducted on the remaining 1,174 tokens.

Prevoicing rates for non-excluded voiced stops revealed clear differences across groups and languages. Arabic monolinguals exhibited a high rate of prevoicing in Arabic /b/ (0.93), whereas English monolinguals showed low prevoicing rates for English /b/ (0.14), and Spanish monolinguals exhibited intermediate rates for Spanish /b/ (0.26). Arabic–English bilinguals showed substantial prevoicing in both Arabic /b/ (0.69) and English /b/ (0.76), alongside moderately high prevoicing in Spanish /b/ (0.62). These patterns provided an empirical basis for examining L2-to-L1 phonetic drift and cross-language interaction in voiced stop production.

3.6 Statistical analysis

All statistical analyses were conducted using IBM SPSS Statistics (Version 28; IBM, Armonk, New York, USA). VOT (ms) was analyzed using linear mixed-effects models with group, condition, and their interaction as fixed effects. Participant was specified as a random intercept to account for repeated measures. Because each condition contained a single lexical item, item-level random effects were not included, and inference was restricted to participant-level generalization. Within the Arabic–English–Spanish group, additional models included Eng_z, Spa_z, and balance as continuous predictors interacting with condition to evaluate proficiency and dominance effects. Prevoicing presence (0/1) was analyzed using generalized linear mixed-effects models with a binomial distribution and logit link, with group and condition as fixed effects and participant as a random intercept. Prevoicing duration was analyzed descriptively and used in supplementary analyses restricted to tokens exhibiting prevoicing. Model assumptions were evaluated using residual diagnostics. All tests were two-tailed with alpha = 0.05. Effect sizes, fixed-effect estimates, and confidence intervals were reported for all models.

4 Results

All results below were computed from the provided dataset after excluding tokens flagged as excluded (26/1,200 tokens; 2.17%), leaving 1,174 analyzable tokens. Descriptive statistics are reported for all conditions and groups, followed by model-based tests aligned with the study's research questions.

4.1 Descriptive statistics for VOT

Table 1 reports VOT descriptives (ms) by group × condition for non-excluded tokens. As expected, English aspirated /p/ (Eng_p_asp) showed long-lag VOT across English-speaking groups, English /p/ after /s/ (Eng_p_after_s) showed short-lag VOT, and Spanish /p/ (Spa_p) showed short-lag VOT. Voiced-stop VOT values were typically negative where closure voicing preceded release, with clear cross-group differences for /b/ categories (Table 1). Figure 1 visualizes voiceless-stop VOT distributions.

Table 1. Descriptive statistics for VOT (ms) by group and condition (non-excluded tokens).

GroupConditionnMean (ms)SDMinMax
Arabic–EnglishArabic /b/74−44.0838.55−104.8813.34
Arabic–EnglishEnglish /b/73−53.9135.13−126.2814.93
Arabic–EnglishEnglish /p/ after /s/7214.645.38−3.2424.11
Arabic–EnglishEnglish /p/ aspirated7471.859.3251.0890.35
Arabic–English–SpanishArabic /b/74−43.8531.92−105.6314.35
Arabic–English–SpanishEnglish /b/75−37.1932.64−93.9317.66
Arabic–English–SpanishEnglish /p/ after /s/7313.844.58−0.2422.01
Arabic–English–SpanishEnglish /p/ aspirated7368.1810.0746.1987.98
Arabic–English–SpanishSpanish /b/74−33.7537.93−106.7621.55
Arabic–English–SpanishSpanish /p/7515.105.223.2828.55
Arabic monolingualArabic /b/73−85.5528.25−127.767.20
English monolingualEnglish /b/720.7417.99−60.5316.94
English monolingualEnglish /p/ after /s/7314.214.332.8022.75
English monolingualEnglish /p/ aspirated7370.9010.3346.1787.26
Spanish monolingualSpanish /b/72−3.8423.26−62.6520.35
Spanish monolingualSpanish /p/7414.174.513.7926.66
Voiceless-stop VOT distributions.
Figure 1. Voiceless-stop VOT distributions.

The Spanish monolingual baseline for Spanish /b/ in this sample was only modestly negative (M = −3.84 ms), which is less prevoiced than many canonical Castilian descriptions. The present discussion therefore treats this group as the empirical control baseline for the current dataset rather than as a universal Spanish norm.

4.2 English voiceless stops: aspiration contrast and s-cluster /p/

To test whether speakers differentiated the English aspiration contrast (Eng_p_asp vs. Eng_p_after_s) and whether learner groups differed from English monolinguals, VOT for English voiceless stops was modeled using a linear mixed model with fixed effects of group, condition, and their interaction, and a participant random intercept.

The English aspiration contrast was robust across groups. Relative to Eng_p_after_s, Eng_p_asp was longer by B = 57.210 ms, 95% CI [54.698, 59.722], p < 0.001 (Table 2). There was no evidence that the Arabic–English–Spanish group differed from the Arabic–English group in the s-cluster baseline (Eng_p_after_s), and the interaction terms did not indicate a reliable reduction of aspiration in Eng_p_asp for the multilingual group relative to the Arabic–English group (Table 2).

Table 2. Linear mixed model for English voiceless-stop VOT (ms).

TermEstimateSEzp95% CI low95% CI high
Intercept14.6430.91216.048012.85516.432
C(Group)[T.Arabic_Eng_Spa]−0.8021.286−0.6230.533−3.3231.719
C(Group)[T.English_Mono]−0.4381.286−0.3400.734−2.9582.083
C(Condition)[T.Eng_p_asp]57.2101.28244.637054.69859.722
C(Group)[T.Arabic_Eng_Spa]:C(Condition)[T.Eng_p_asp]−2.8691.813−1.5830.113−6.4230.684
C(Group)[T.English_Mono]:C(Condition)[T.Eng_p_asp]−0.5141.813−0.2840.777−4.0673.039

Data: Eng_p_after_s and Eng_p_asp; groups = Arabic–English (reference), Arabic–English–Spanish, English monolingual. Reference levels: Group = Arabic–English; Condition = Eng_p_after_s. Random effect: participant intercept (estimated variance approximately 0 for this model in these data).

4.3 Spanish /p/: cross-language coupling test

Spanish /p/ VOT was compared between the Arabic–English–Spanish group and Spanish monolinguals using a linear mixed model with group as a fixed effect and participant random intercept. Spanish monolinguals did not differ reliably from Arabic–English–Spanish speakers in Spanish /p/ VOT (B = −0.933 ms, 95% CI [−2.569, 0.703], p = 0.264). Descriptively, both groups produced short-lag VOT (Table 1). This pattern did not support systematic over-aspiration of Spanish /p/ among multilinguals in this dataset.

4.4 Dominance effects within the Arabic–English–Spanish group

Within the Arabic–English–Spanish group, VOT for Eng_p_after_s, Eng_p_asp, and Spa_p was modeled as a function of condition, balance (Eng_z - Spa_z), and their interaction, with a participant random intercept. Balance did not reliably predict VOT, and there were no reliable balance × condition interactions (Table 3). Dominance variation, as operationalized here, did not explain additional variance in voiceless-stop VOT in the multilingual group.

Table 3. Linear mixed model within the Arabic–English–Spanish group: VOT predicted by condition, balance, and interactions.

TermEstimateSEzp95% CI low95% CI high
Intercept13.8490.83616.572012.21115.487
C(Condition)[T.Eng_p_asp]54.3041.13647.820052.07956.530
C(Condition)[T.Spa_p]1.2521.1281.1100.267−0.9583.462
Balance_Eng_minus_Spa0.2030.6150.3300.741−1.0011.408
C(Condition)[T.Eng_p_asp]:Balance_Eng_minus_Spa0.6250.8470.7380.461−1.0362.286
C(Condition)[T.Spa_p]:Balance_Eng_minus_Spa0.4350.8340.5220.602−1.1992.070

Data: Eng_p_after_s (reference), Eng_p_asp, Spa_p. Predictor: Balance = Eng_z – Spa_z. Random effect: participant intercept.

4.5 Voiced-stop VOT: arabic /b/ drift and cross-language comparisons

Given the structural missingness across groups, not all groups produced all voiced-stop conditions, so voiced-stop VOT was tested with targeted models aligned with the research questions.

4.5.1 Arabic /b/ (drift-relevant comparison)

Arabic /b/ VOT differed strongly between Arabic monolinguals and the Arabic L1 learner groups. Compared with the Arabic–English group (reference), Arabic monolinguals showed substantially more negative VOT [B = −41.484 ms, 95% CI (−52.325, −30.644), p < 0.001], indicating longer prevoicing lead. The Arabic–English–Spanish group did not differ from the Arabic–English group in Arabic /b/ VOT (Table 4).

Table 4. Linear mixed model for voiced-stop VOT (ms) by language context.

TermEstimateSEzp95% CI low95% CI high
Intercept−44.0683.898−11.3050−51.709−36.428
C(Group)[T.Arabic_Eng_Spa]0.2245.5130.0410.968−10.58111.028
C(Group)[T.Arabic_Mono]−41.4845.531−7.5000−52.325−30.644

Arabic /b/ (condition = Arb_b). Reference: Arabic–English.

4.5.2 English /b/

For English /b/, English monolinguals showed significantly less negative VOT than the Arabic-English group [B = 54.648 ms, 95% CI [45.063, 64.232], p < 0.001]. The Arabic–English–Spanish group also differed from the Arabic-English group [B = 16.720 ms, 95% CI (7.233, 26.206), p = 0.001], reflecting less negative English /b/ VOT among multilinguals than among Arabic–English bilinguals (Table 5).

Table 5. English /b/ (condition = Eng_b).

TermEstimateSEzp95% CI low95% CI high
Intercept−53.9113.446−15.6460−60.665−47.158
C(Group)[T.Arabic_Eng_Spa]16.7204.8403.4540.0017.23326.206
C(Group)[T.English_Mono]54.6484.89011.175045.06364.232

Reference: Arabic–English.

4.5.3 Spanish /b/

Spanish monolinguals showed significantly less negative Spanish /b/ VOT than the Arabic-English-Spanish group [B = 29.909 ms, 95% CI (19.739, 40.079), p < 0.001], consistent with more lead voicing among the multilingual speakers in Spanish /b/ (Table 6). Each panel was estimated as a separate model with a participant random intercept, and the relevant reference group is indicated in the table caption.

Table 6. Spanish /b/ (condition = Spa_b).

TermEstimateSEzp95% CI low95% CI high
Intercept−33.7543.645−9.2590−40.899−26.609
C(Group)[T.Spanish_Mono]29.9095.1895.764019.73940.079

4.6 Prevoicing presence (0/1): group differences by language context

Prevoicing presence was evaluated using clustered logistic modeling with participant-level clustering (robust standard errors). Odds ratios (ORs) and 95% confidence intervals are reported in Tables 7–9, alongside the observed prevoicing proportions summarized above.

Table 7. Arabic /b/ prevoicing (condition = Arb_b).

TermLog-odds (B)SEzpOR95% CI OR low95% CI OR high
Intercept0.7960.2183.64702.2181.4453.403
C(Group)[T.Arabic_Eng_Spa]0.5730.4011.4290.1531.7740.8083.896
C(Group)[T.Arabic_Mono]1.8130.4803.78106.1302.39515.691

4.6.1 Arabic /b/

Arabic monolinguals were significantly more likely to prevoice Arabic /b/ than the Arabic–English group [OR = 6.133, 95% CI (2.396, 15.699), p < 0.001]. The Arabic–English–Spanish group did not differ reliably from the Arabic–English group (Table 7).

4.6.2 English /b/

English monolinguals were dramatically less likely to prevoice English /b/ than the Arabic–English group [OR = 0.035, 95% CI (0.015, 0.083), p < 0.001]. The Arabic–English–Spanish group did not differ reliably from the Arabic–English group in prevoicing likelihood for English /b/ (Table 8).

Table 8. English /b/ prevoicing (condition = Eng_b).

TermLog-odds (B)SEzpOR95% CI OR low95% CI OR high
Intercept1.5300.3005.09604.6162.5628.312
C(Group)[T.Arabic_Eng_Spa]−0.3770.413−0.9130.3610.6860.3051.540
C(Group)[T.English_Mono]−3.3540.441−7.61100.0350.0150.083

Reference: Arabic–English.

4.6.3 Spanish /b/

Spanish monolinguals were less likely to prevoice Spanish /b/ than the Arabic–English–Spanish group [OR = 0.219, 95% CI (0.116, 0.415), p < 0.001], indicating a higher likelihood of prevoicing among the multilingual speakers in Spanish /b/ (Table 9).

Table 9. Spanish /b/ prevoicing (condition = Spa_b).

TermLog-odds (B)SEzpOR95% CI OR low95% CI OR high
Intercept0.5040.1932.6190.0091.6561.1352.415
C(Group)[T.Spanish_Mono]−1.5180.325−4.66600.2190.1160.415

Reference: Arabic–English–Spanish.

5 Discussion

5.1 Overview and theoretical framing

The present study investigated VOT in Arabic L1 learners of English and Spanish within the framework of the speech learning model-revised, with the aim of examining how multiple phonetic systems are established, maintained, and reshaped in adult multilingual speakers. By combining controlled production data with mixed-effects modeling and continuous measures of proficiency and dominance, the study provided a stringent test of SLM-r predictions concerning gradient learning, shared phonetic representations, cross-language interaction, and L2-to-L1 phonetic drift (Flege and Bohn, 2021).

The pattern that emerged is asymmetrical across phonetic dimensions. Voiceless stops showed strong language-specific stability, whereas voiced stops showed clear evidence of cross-language interaction. This asymmetry is theoretically informative because it suggests that not all categories are equally vulnerable to multilingual influence. Categories tied to robust aspiration contrasts may remain relatively stable once established, whereas lead-voicing patterns may be more susceptible to interaction, drift, and partial restructuring.

5.2 Maintenance of the English aspiration contrast

A central finding of the study is the robust maintenance of the English aspiration contrast across all groups producing English. English aspirated /p/ exhibited long-lag VOT values that were reliably distinct from English /p/ in s-clusters, and this contrast was preserved by both the Arabic–English bilinguals and the Arabic–English–Spanish multilinguals. The absence of group differences or interaction effects indicates that learners successfully implemented the English laryngeal timing contrast even in the presence of an additional non-native language.

This outcome aligns with research showing that adult learners can acquire and stabilize new phonetic targets when contrasts are perceptually salient and consistently reinforced by input and use (Cho et al., 2019; Flege, 1991; Munro et al., 2024). Within SLM-r, the result supports the claim that phonetic learning remains available across the lifespan and that new categories can be formed without necessarily destabilizing existing ones (Flege and Bohn, 2021).

5.3 Spanish /p/ and constraints on cross-L2 phonetic interaction

Spanish /p/ produced by the Arabic–English–Spanish group did not show evidence of systematic over-aspiration relative to Spanish monolinguals. Instead, Spanish /p/ clustered in the short-lag range, remained clearly distinct from English aspirated /p/, and closely approximated the Spanish monolingual baseline in this dataset. This finding is compatible with multilingual work showing that speakers can maintain language-specific VOT targets even in shared phonetic space (Antoniou et al., 2010; Geiss et al., 2022; Souganidis et al., 2024; Stoehr et al., 2024).

At the same time, the present data also permit a more cautious interpretation than simple category separation alone. Because Arabic voiceless stops are commonly short-lag, the target-like realization of Spanish /p/ may also reflect facilitative transfer from Arabic timing patterns rather than only resistance to English aspiration. Early Arabic–English work and later bilingual studies show that phonetic transfer can be asymmetric and context dependent (Antoniou et al., 2011; Flege and Port, 1981; Olson, 2013). Without an Arabic voiceless bilabial comparison in the present dataset, it is not possible to disentangle completely whether Spanish /p/ reflects newly stabilized Spanish-specific control, facilitative transfer from Arabic, or both. This is an important limitation rather than a contradiction of the main result.

5.4 Proficiency, dominance, and gradient variation

The study explicitly modeled proficiency and relative dominance as continuous predictors. Within the Arabic–English–Spanish group, relative dominance between English and Spanish did not reliably predict VOT variation for voiceless stops. This null effect does not challenge gradient learning accounts. It suggests instead that once the English aspiration contrast and the Spanish short-lag target are established in a controlled task, additional variation in relative dominance may not substantially modulate these particular phonetic outcomes.

This interpretation is consistent with evidence that dominance effects are feature specific and may emerge more clearly for categories that are less stable, less frequent, or more weakly differentiated perceptually (Souganidis et al., 2024; Takahashi, 2023). In the present dataset, the voiceless-stop categories appear to have been sufficiently entrenched to resist measurable modulation by the Balance index.

5.5 Voiced stops and evidence for L2-to-L1 phonetic drift

The clearest evidence for cross-language influence emerged in voiced-stop production. Arabic monolinguals exhibited robust prevoicing and strongly negative VOT values for Arabic /b/, consistent with prior descriptions of Arabic stop systems (Al-Tamimi et al., 2021; Lisker and Abramson, 1964). In contrast, the Arabic–English and Arabic–English–Spanish groups both showed substantially reduced lead voicing in Arabic /b/, indicating a shift away from monolingual Arabic norms.

This pattern is consistent with L2-to-L1 phonetic drift, a phenomenon documented in bilingual research and predicted by SLM-r (Chang, 2012, 2013, 2019; Flege and Bohn, 2021). The absence of a difference between the two Arabic L1 learner groups indicates that adding Spanish did not measurably attenuate or amplify the Arabic /b/ pattern in these data. Because Spanish voiced stops are commonly described as strongly prevoiced in benchmark studies (Flege and Eefting, 1987; Rosner et al., 2000), the observed drift pattern is unlikely to be attributable to Spanish alone and is more plausibly linked to English exposure or to multilingual experience more broadly than to Spanish-specific influence.

The English and Spanish voiced-stop results further underscore the bidirectional nature of phonetic interaction. Arabic L1 learners exhibited higher rates of prevoicing in English /b/ and Spanish /b/ than the monolingual controls for those languages, which points to persistence of Arabic voicing patterns. At the same time, the learner groups did not simply reproduce Arabic monolingual values in either language. The result is best characterized as partial convergence rather than wholesale transfer, consistent with SLM-r and with prior accounts of bilingual stop production that allow coexistence of shared phonetic pressure and language-specific implementation (Antoniou et al., 2010, 2011; Olson, 2013).

Reviewer 2's concern about the Spanish /b/ baseline is well taken. Many descriptions of Castilian Spanish report considerably more negative VOT values than the −3.84 ms mean observed here (Flege and Eefting, 1987; Lisker and Abramson, 1964; Rosner et al., 2000). Because detailed hometown and current-residence metadata were not retained for all monolingual controls, the Spanish monolingual group in the present study should be treated as the relevant empirical baseline for this dataset, not as a universal benchmark for all varieties of Spanish. The mismatch with canonical reports should therefore be read as a limitation on cross-study comparison rather than as evidence against the current within-dataset pattern.

5.6 Implications for the speech learning model-revised

The findings support several core claims of the speech learning model-revised. Adult multilingual speakers maintained distinct language-specific targets for the English aspiration contrast and for Spanish short-lag /p/, which is consistent with category formation and stability in shared phonetic space. The same dataset also revealed substantial interaction in voiced-stop production, including reduced Arabic /b/ prevoicing among the Arabic L1 learner groups and elevated prevoicing in English /b/ and Spanish /b/ relative to the respective monolingual controls. These results are compatible with the SLM-r view that phonetic systems remain plastic and mutually influential across the lifespan.

The asymmetry between voiceless and voiced stops is especially informative. It suggests that multilingual phonetic interaction is feature selective rather than uniform across the stop system. Aspirated vs. short-lag voiceless contrasts may be sufficiently salient and stable to resist measurable drift in the present task, whereas lead-voicing patterns appear more vulnerable to restructuring and partial carryover across languages. Future theorizing about multilingual stop systems may therefore benefit from feature-specific accounts of stability and interaction rather than expecting parallel behavior across all stop categories.

6 Limitations and future directions

Several limitations should be acknowledged. First, the study used a controlled reading task with a restricted set of lexical items and only one target word per condition. This design supported tight phonetic control but limits lexical generalization and leaves open the possibility that different lexical items or more spontaneous tasks would show somewhat different VOT distributions.

Second, the absence of an Arabic voiceless bilabial comparison item limits the interpretation of Spanish /p/. Although the multilingual group clearly avoided English-like over-aspiration in Spanish, the present design cannot fully disentangle stable Spanish category control from facilitative transfer from Arabic short-lag voiceless timing. Future work should include a broader Arabic voiceless-stop comparison set to test this distinction more directly.

Third, detailed hometown-level provenance and current-residence information were not retained for all monolingual controls. This limitation is especially important for interpreting Spanish /b/, because benchmark studies frequently report more robust prevoicing than was observed in the present Spanish monolingual sample. Subsequent work should document dialect background, bilingual context, and language-contact history more systematically for all comparison groups.

Fourth, the design is cross-sectional. The observed differences between Arabic monolinguals and the Arabic L1 learner groups are consistent with phonetic drift, but they do not by themselves track the temporal dynamics of change. Longitudinal designs would provide stronger evidence about when Arabic /b/ begins to shift, whether the shift stabilizes, and whether it is reversible under changing language-use conditions.

Future work should also integrate perception and production more directly. Because SLM-r treats phonetic learning as grounded in cross-language category relations, a combined perception-production design would clarify whether the voiced-stop patterns reported here reflect changes in articulatory implementation alone or deeper reorganization of multilingual phonetic categories.

7 Conclusion

This study provides a multilingual account of voice onset time in Arabic L1 speakers learning English and Spanish. The results show stable maintenance of the English aspiration contrast and short-lag Spanish /p/, alongside substantial cross-language interaction in voiced-stop production. Arabic /b/ was less prevoiced in both Arabic L1 learner groups than in Arabic monolinguals, while English /b/ and Spanish /b/ showed more prevoicing in the learner groups than in the corresponding monolingual controls.

These findings support key claims of the speech learning model-revised while also pointing to a more nuanced interpretation of multilingual stop production. Shared phonetic space did not eliminate language-specific control for voiceless stops, but it did coincide with clear interaction and drift in voiced-stop timing. The study therefore contributes to a growing body of evidence that adult multilingual phonetic systems remain dynamic, selectively interactive, and sensitive to the structure of particular phonetic contrasts.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Junjie Wu, Tianjin Normal University, China

Reviewed by: Dalia Abdulkader, King Saud University, Saudi Arabia

Simona Montanari, California State University, Los Angeles, United States

Statements

Conflict of interest

The author(s) declared that this work was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Funding

The author(s) declared that financial support was not received for this work and/or its publication.

Data availability

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

Ethics approval

The studies involving humans were approved by University of Jordan Ethics Committee. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study. Written informed consent was obtained from the individual(s) for the publication of any potentially identifiable images or data included in this article.

Author contributions

HA: Writing – review & editing, Writing – original draft. MA-D: Writing – original draft, Writing – review & editing. RA: Writing – review & editing, Writing – original draft. MA: Writing – original draft, Writing – review & editing.

References

  1. Aldamen H., Al-Deaibes M. (2023a). Arabic emphatic consonants as produced by English speakers: an acoustic study. Heliyon 9: e13401. doi: doi:10.1016/j.heliyon.2023.e13401
  2. Aldamen H., Al-Deaibes M. (2023b). Perception and production of L2 Arabic emphatic consonants: the role of communicative and traditional form-based approaches. Ampersand 10: 100105. doi: doi:10.1016/j.amper.2022.100105
  3. Al-Damen H., Almashour M., Al-Deaibes M., Alsharefeen R. (2025). Voice onset time in the production of English stops: a comparative study of native and Arabic-speaking EFL learners. Forum Linguist. Stud. 7, 747 – 757. doi: doi:10.30564/fls.v7i7.9466
  4. Al-Deaibes M., Mashaqba B., Huneety A., Guba M. N. A. (2026). Effects of gender and age on voice onset time in rural Jordanian Arabic. J. Psycholinguist. Res. 55: 3. doi: doi:10.1007/s10936-025-10176-y
  5. Alharbi A. S., Foltz A., Mennen I. (2023). L2 acquisition and L1 attrition of VOTs of voiceless plosives in highly proficient late bilinguals. Second Lang. Res. 39, 1085 – 1111. doi: doi:10.1177/02676583221124011
  6. Al-Tamimi F., Tarawneh R., Howell P. (2021). Development of voice onset time in Arabic. Lingua 262: 103117. doi: doi:10.1016/j.lingua.2021.103117
  7. Antoniou M., Best C. T., Tyler M. D., Kroos C. (2010). Language context elicits native-like stop voicing in early bilinguals' productions in both L1 and L2. J. Phon. 38, 640 – 653. doi: doi:10.1016/j.wocn.2010.09.005
  8. Antoniou M., Best C. T., Tyler M. D., Kroos C. (2011). Inter-language interference in VOT production by L2-dominant bilinguals: asymmetries in phonetic code-switching. J. Phon. 39, 558 – 570. doi: doi:10.1016/j.wocn.2011.03.001
  9. Chang C. B. (2012). Rapid and multifaceted effects of second-language learning on first-language speech production. J. Phon. 40, 249 – 268. doi: doi:10.1016/j.wocn.2011.10.007
  10. Chang C. B. (2013). A novelty effect in phonetic drift of the native language. J. Phon. 41, 520 – 533. doi: doi:10.1016/j.wocn.2013.09.006
  11. Chang C. B. (2019). Language change and linguistic inquiry in a world of multicompetence: sustained phonetic drift and its implications for behavioral linguistic research. J. Phon. 74, 96 – 113. doi: doi:10.1016/j.wocn.2019.03.001
  12. Cho T., Whalen D. H., Docherty G. (2019). Voice onset time and beyond: exploring laryngeal contrast in 19 languages. J. Phon. 72, 52 – 65. doi: doi:10.1016/j.wocn.2018.11.002
  13. Davis K. (1995). Phonetic and phonological contrasts in the acquisition of voicing: voice onset time production in Hindi and English. J. Child Lang. 22, 275 – 305. doi: doi:10.1017/S030500090000979X
  14. Face T. L., Menke M. R. (2020). L2 acquisition of Spanish VOT by English-speaking immigrants in Spain. Stud. Hisp. Lusoph. Linguist. 13, 361 – 389. doi: doi:10.1515/shll-2020-2034
  15. Flege J. E. (1991). Age of learning affects the authenticity of voice-onset time (VOT) in stop consonants produced in a second language. J. Acoust. Soc. Am. 89, 395 – 411. doi: doi:10.1121/1.400473
  16. Flege J. E., Bohn O.-S. (2021). “The revised speech learning model (SLM-r),” in Second Language Speech Learning: Theoretical and Empirical Progress, ed. R. Wayland (Cambridge: Cambridge University Press), 3 – 83. doi: doi:10.1017/9781108886901.002
  17. Flege J. E., Eefting W. (1987). Production and perception of English stops by native Spanish speakers. J. Phon. 15, 67 – 83. doi: doi:10.1016/S0095-4470(19)30538-8
  18. Flege J. E., Port R. (1981). Cross-language phonetic interference: arabic to English. Lang. Speech 24, 125 – 146. doi: doi:10.1177/002383098102400202
  19. Geiss M., Gumbsheimer S., Lloyd-Smith A., Schmid S., Kupisch T. (2022). Voice onset time in multilingual speakers: Italian heritage speakers in Germany with L3 English. Stud. Second Lang. Acquis. 44, 435 – 459. doi: doi:10.1017/S0272263121000280
  20. Goldrick M., Shrem Y., Kilbourn-Ceron O., Baus C., Keshet J. (2021). Using automated acoustic analysis to explore the link between planning and articulation in second language speech production. Lang. Cogn. Neurosci. 36, 824 – 839. doi: doi:10.1080/23273798.2020.1805118
  21. Hamzah M. H., Madbouly A. E. S., Halim H. A., Abdullah A. H. (2020). The production of the English stop voicing contrast by Arab L2 speakers of English. Indones. J. Appl. Linguist. 10, 522 – 531. doi: doi:10.17509/ijal.v10i2.28615
  22. Kehoe M., Kannathasan K. (2021). Development of voice onset time in monolingual and bilingual French-speaking children. Lingua 252: 102937. doi: doi:10.1016/j.lingua.2020.102937
  23. Kelly N. E., El Houry M., Ghamloush F. (2021). Monolingual vs. codeswitching modes: voice onset time in Lebanese Arabic and English. Proc. Meet. Acoust. 42: 060012. doi: doi:10.1121/2.0001422
  24. Kornder L., Mennen I. (2021). Longitudinal developments in bilingual second language acquisition and first language attrition of speech: the case of Arnold Schwarzenegger. Languages 6: 61. doi: doi:10.3390/languages6020061
  25. Lisker L., Abramson A. S. (1964). A cross-language study of voicing in initial stops: acoustical measurements. Word 20, 384 – 422. doi: doi:10.1080/00437956.1964.11659830
  26. Lowenstein J. H., Nittrouer S. (2008). Patterns of acquisition of native voice onset time in English-learning children. J. Acoust. Soc. Am. 124, 1180 – 1191. doi: doi:10.1121/1.2945118
  27. Macken M. A., Barton D. (1980). The acquisition of the voicing contrast in English: a study of voice onset time in word-initial stop consonants. J. Child Lang. 7, 41 – 74. doi: doi:10.1017/S0305000900007029
  28. Morales-Front A. (2018). “Voice onset time in advanced SLA,” in The Handbook of Advanced Proficiency in Second Language Acquisition, eds. L. Ortega, and S. Han (Malden, MA: Wiley-Blackwell), 223 – 244. doi: doi:10.1002/9781119261650.ch17
  29. Munro M., Derwing T. M., Saito K. (2024). Aspiring to aspirate: L2 acquisition of English word-initial /p/ over 10 years. Lang. Speech 68, 247 – 261. doi: doi:10.1177/00238309241264296
  30. Nagle C. (2019). A longitudinal study of voice onset time development in L2 Spanish stops. Appl. Linguist. 40, 86 – 107. doi: doi:10.1093/applin/amx011
  31. Olson D. J. (2013). Bilingual language switching and selection at the phonetic level: asymmetrical transfer in VOT production. J. Phon. 41, 407 – 420. doi: doi:10.1016/j.wocn.2013.07.005
  32. Olson D. J. (2019). Feature acquisition in second language phonetic development: evidence from phonetic training. Lang. Learn. 69, 366 – 404. doi: doi:10.1111/lang.12336
  33. Olson D. J. (2021). Phonetic feature size in second language acquisition: examining VOT in voiceless and voiced stops. Second Lang. Res. 38, 913 – 940. doi: doi:10.1177/02676583211008951
  34. Rosner B. S., López-Bascuas L. E., García-Albea J. E., Fahey R. P. (2000). Voice-onset times for Castilian Spanish initial stops. J. Phon. 28, 217 – 224. doi: doi:10.1006/jpho.2000.0113
  35. Souganidis C., Molinaro N., Stoehr A. (2024). Bilinguals produce language-specific voice onset time in two true-voicing languages: the case of Basque-Spanish early bilinguals. Linguist. Approaches Biling. 14, 370 – 399. doi: doi:10.1075/lab.21081.sou
  36. Stoehr A., Benders T., van Hell J. G., Fikkert P. (2017). Second language attainment and first language attrition: the case of VOT in immersed Dutch-German late bilinguals. Second Lang. Res. 33, 483 – 518. doi: doi:10.1177/0267658317704261
  37. Stoehr A., Jevtović M., de Bruin A., Martin C. D. (2024). Phonetic and lexical crosslinguistic influence in early Spanish-Basque-English trilinguals. Lang. Learn. 74, 332 – 364. doi: doi:10.1111/lang.12598
  38. Takahashi C. (2023). L1 vowel perceptual boundary shift as a result of L2 vowel learning. J. Phon. 100: 101265. doi: doi:10.1016/j.wocn.2023.101265
  39. Zen E. L. (2020). Role of regional language background and speech styles on the production of voice onset time (VOT) in English among Indonesian multilinguals. Indones. J. Appl. Linguist. 10, 359 – 368. doi: doi:10.17509/ijal.v10i2.28604

Citation

Aldamen, H., et al. (2026). Voice onset time in Arabic L1 learners of English and Spanish: evidence for stability, interaction, and phonetic drift in multilingual speech. Frontiers in Language Sciences, Volume 5. https://doi.org/
Received
Accepted
Published
Volume
5 - 2026

Copyright

Copyright © 2026 Aldamen, Al-Deaibes, Alsharefeen and Almashour.

License: https://creativecommons.org/licenses/by/4.0/

*Correspondence: Mutasim Al-Deaibes

Article metrics

Views
0
Downloads
0
Citations
0
Shares
0

Related articles

Original ResearchDo large language models keep the hedge? Uncertainty preservation when large language models summarize aviation and maritime incident reportsSimon BaradziejOriginal ResearchMetaphors' journeys across time and genre: tracking the evolution of literary metaphors with temporal embeddingsVeronica Mangiaterra, Chiara Barattieri di San Pietro, Paolo Canal, Valentina BambiniConceptual AnalysisBeyond the bridge: an integrative developmental framework for first-language reading fluencyJános Steklács