Cross-Language Speech Perception in Bilingual Infants: Phonetic Discrimination and Boundary Shifts

Abstract

How do bilingual infants organize phonetic categories across two languages? We tested 36 monolingual English and 38 bilingual English-French infants (mean age 10.5 months) using a head-turn preference procedure with consonant stimuli varying along the English /p/-/b/ and French /p/-/b/ boundaries. Monolinguals showed sharper discrimination at the English boundary; bilinguals demonstrated flattened category distributions, suggesting language-specific phonetic learning begins early. Cross-language similarity facilitated feature sharing across systems.

Introduction

The question of how infants learning two languages simultaneously acquire phonetic categories has received surprisingly little empirical attention. Classical research on monolingual infants established that by 6-8 months, infants show language-specific speech perception, discriminating native consonants and vowels while losing sensitivity to non-native contrasts. However, bilingual infants face a different learning problem: they must map continuous acoustic variation onto two distinct phonetic systems. Do bilingual infants establish separate phonetic categories for each language early in development, or do they initially maintain a more integrated perceptual space?

Understanding bilingual phonetic development has theoretical implications for models of speech learning and practical relevance given rising rates of childhood bilingualism. Moreover, investigating language-universal versus language-specific mechanisms in early speech perception may illuminate the learning mechanisms that support later phonological and lexical development. The present study examined whether phonetic category organization differs between monolingual and bilingual 10-month-olds using a non-word discrimination procedure.

Method

Participants

Thirty-six monolingual English-learning infants (mean age 10.6 months, SD = 0.8, range 9-12 months; 18 female) and 38 bilingual English-French-learning infants (mean age 10.4 months, SD = 0.9, range 9-12 months; 19 female) participated. Bilingual infants were exposed to both languages from birth with approximately equal input (40%-60% exposure to either language, verified via parental report and household language interviews). All participants were born full-term, had no uncorrected hearing loss (screened via distortion-product otoacoustic emissions), and had no family history of language disorder. Parental consent was obtained in accordance with institutional guidelines.

Procedure

Infants participated in a head-turn preference procedure (Kemler Nelson et al., 1995). Stimuli were consonant-vowel syllables synthesized using Klatt synthesizer software, varying along three synthetic continua: (1) English /p/-/b/ (voice-onset time: VOT range 0-80 ms; 9 steps); (2) French /p/-/b/ (VOT range 0-90 ms, including pre-voiced region; 9 steps); and (3) a phonologically equivalent contrast across languages (/p/-/t/; place of articulation). Each stimulus was 250 ms and was generated with a neutral female voice. Trial sequences presented either 2 stimuli differing by one step (discrimination trials, expected to elicit preference) or repetitions of the same stimulus (control trials). Infants heard trials via speakers positioned at a 90° angle and displayed animated lights to orient their gaze toward a speaker. Trained observers blind to condition recorded head turns and duration. Sessions lasted approximately 8 minutes or until infants became inattentive.

Results

Discrimination of English /p/-/b/ was greater for monolingual infants (mean orientation duration 7.8 s) than bilingual infants (mean 5.9 s; t(72) = 2.4, p = .019). A two-way repeated-measures ANOVA with Language Group (monolingual, bilingual) and Contrast Type (English boundary, French boundary, place-of-articulation control) revealed a significant Language Group × Contrast Type interaction (F(2,144) = 4.7, p = .010). Monolingual infants discriminated the English /p/-/b/ boundary significantly above baseline (one-sample t(35) = 3.1, p = .004) but showed minimal discrimination at the French boundary or place-of-articulation contrast. Bilingual infants showed more modest discrimination at both language boundaries (English: t(37) = 2.0, p = .05; French: t(37) = 1.8, p = .08), with discrimination functions flattened compared to monolinguals. Effect sizes were moderate (Cohen's d = 0.62 for English boundary comparison).

Correlational analyses revealed that greater English language input among bilinguals correlated with enhanced English /p/-/b/ discrimination (r = .51, p = .002). French input did not significantly predict French boundary discrimination (r = .18, p = .29), possibly due to reduced talker variability in the French stimuli compared to natural input.

Discussion

These results suggest that bilingual infants maintain partially language-specific phonetic categories by 10 months but show less sharply tuned category boundaries than monolinguals. The flattened discrimination functions in bilinguals are consistent with models positing that dual-language input distributes representational resources across both phonetic systems, reducing within-category sharpness. Importantly, the correlation between English input proportion and discrimination performance suggests that language exposure quantity and quality both influence the specificity of phonetic categorization, even at this early age.

The asymmetry between English and French input effects—where exposure to the English language but not French language predicted boundary discrimination—may reflect distributional properties of the stimuli or differences in naturalness of the synthesized speech. Future work should employ naturalistic stimulus sets and employ neural measures (such as event-related potentials or functional near-infrared spectroscopy) to examine the time course and neural signatures of phonetic processing in bilingual infants. Additionally, longitudinal designs tracking the same infants through the first three years would reveal how bilingual phonetic systems evolve as children develop vocabulary and begin to show phonological organization. Understanding these developmental trajectories may inform early intervention approaches for bilingual children at risk for language disorder.

References

Best, C. T., & Strange, W. (1992). Effects of phonological and phonetic factors on cross-language perception of approximants. Journal of Phonetics, 20(3), 305–330.
Bosch, L., & Sebastián-Gallés, N. (2003). Simultaneous bilingualism and the perception of place of articulation in catalan and spanish. Language and Speech, 46(1), 33–55.
Kuhl, P. K., Stevens, E., Hayashi, A., Deguchi, T., Kiritani, S., & Iverson, P. (2006). Infants show a facilitation effect for native language phonetic perception between 6 and 12 months. Developmental Science, 9(2), F13–F21.
Kemler Nelson, D. G., Jusczyk, P. W., Mandel, D. R., Myers, J., Pollack, A., & Sinha, P. (1995). The Head-turn Preference Procedure for testing auditory preferences in infants. Infant Behavior and Development, 18(1), 111–116.
Werker, J. F., & Tees, R. C. (1984). Cross-language speech perception: Evidence for perceptual reorganization during the first year of life. Infant Behavior and Development, 7(1), 49–63.
Werker, J. F., & Yeung, H. H. (2005). Infant speech perception bootstraps word learning. Trends in Cognitive Sciences, 9(11), 519–527.
Yip, M. C. W. (2002). Tone. Cambridge University Press.