It is well established that early bilinguals who speak languages that differ in the phonetic implementation of the voicing contrast have language-specific voicing systems. This study investigates voicing separation in bilinguals who speak two true-voicing languages, Basque and Spanish. We also describe the voicing system in Gipuzkoan Spanish and Gipuzkoan Basque, which is the closest dialect to Standard Basque and it has not yet been investigated experimentally. Twenty Basque-Spanish early bilingual speakers of Gipuzkoan dialects participated in two picture naming tasks. We described their voicing system by measuring voice onset time (VOT) in both Gipuzkoan Basque and Spanish, and used linear mixed-effects models to investigate between-language production differences. Our results show for the first time that adult early bilinguals who speak two true-voicing languages produce language-specific VOT in ‘voiced’ plosives. This finding demonstrates that bilinguals’ phonetic systems during production are more fine-grained than previously assumed, and contributes to a deeper understanding of granularity in early bilingual phonetic systems.

Since the COVID-19 pandemic started, conducting experiments online is increasingly common, and face masks are often used in everyday life. It remains unclear whether phonetic detail in speech production is captured adequately when speech is recorded in internet-based experiments or in experiments conducted with face masks. We tested 55 Spanish–Basque–English trilinguals in picture naming tasks in three conditions: online, laboratory-based with surgical face masks, and laboratory-based without face masks (control). We measured plosive voice onset time (VOT) in each language, the formants and duration of English vowels /iː/ and /ɪ/, and the Spanish/Basque vowel space. Across conditions, there were differences between English and Spanish/Basque VOT and in formants and duration between English /iː/–/ɪ/; between conditions, small differences emerged. Relative to the control condition, the Spanish/Basque vowel space was larger in online testing and smaller in the face mask condition. We conclude that testing online or with face masks is suitable for investigating phonetic detail in within-participant designs although the precise measurements may differ from those in traditional laboratory-based research.