Ultrasound imaging of the tongue (US) requires probe stabilization underneath the jaw. Pucher et al. (2020) reported various quality assessments for their Ultrafit (Spreafico et al., 2018) and an earlier metallic headset (Scobbie et al., 2008), compared to naturalistic speech; wearing the latter produced a statistically significant effect on F1 of the /u/ vowel. Here, I present formant data of 9939 monophthongs produced by 13 participants while (not) wearing the 3-D-printed ALPHUS headset (Chen et al., 2024). Wordlist readings were recorded using Audacity or Articulate Assistant Advanced through a Focusrite audio interface with an AT875R shotgun microphone. Recordings were segmented automatically using MFA (McAuliffe et al., 2017) and checked in Praat (Boersma & Weenink, 2024), followed by formant measurements utilizing FastTrack (Barreda, 2021). I then fit linear mixed effects models using lme4 (Bates et al., 2015) in R to determine whether F1, F2, and vowel duration were statistically different dependent on condition (US/no US). The best models included condition and gender as fixed effects (no interaction term), and by-subject, by-word, and by-vowel random slopes for condition. None of the models incorporated any significant main effects (including condition), indicating that data collection utilizing the ALPHUS headset generates naturalistic acoustic data.
Matthias Heyne (2025) studied this question.