Is Feature Timing the Key to Phonological Acquisition? Naturalistic Evidence from EEG Encoding Models across the First Five Years of Life
Poster D38 in Poster Session D with Social Hour, Friday, October 7, 5:30 - 7:15 pm EDT, Millennium Hall
Katharina Menn1,2,3, Claudia Männel2,4, Lars Meyer1,5; 1RG Language Cycles, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 2Department of Neuropsychology, Max Planck Institute for Human Cognitive and Brain Sciences, Leipzig, Germany, 3IMPRS NeuroCom, Leipzig, Germany, 4Department of Audiology and Phoniatrics, Charité – Universitätsmedizin Berlin, Germany, 5Clinic for Phoniatrics and Pedaudiology, University Hospital Münster, Germany
Infants rapidly tune into their native language (Kuhl, 2004): By the age of 6 months, they show enhanced processing of native phonemes (Kuhl et al., 2007; Ortiz-Mantilla et al., 2013). Sensitivity to phonemes presumes the ability to segment the corresponding acoustic segments in speech, which are as short as ~50 ms (Leong & Goswami, 2015). This is a challenge for the newborn brain, where slow electrophysiological activity prevails (Anderson & Perone, 2018) that only offers long temporal receptive windows (Hochmann & Kouider, 2022). How do infants acquire a native phoneme inventory at an early age despite their initial electrophysiological slowness? A high phoneme rate does not mean that all individual phoneme features change at a high rate. In speech, some features (e.g., voicing, place of articulation) extend over sequences of multiple adjacent phonemes. This results in feature-continuous stretches that would fit infants’ long temporal receptive windows. We thus hypothesized that feature duration explains the early acquisition of native phoneme features in the absence of fast electrophysiological activity. We recorded the electroencephalogram (EEG) from n = 77 children aged 0;3–4;6 years. Children heard translation-equivalent stories in their native language (German) and an unfamiliar language (French). We quantified the processing of phonological features through the prediction accuracy of EEG encoding models (temporal response functions, TRFs). We compared the prediction accuracy of the TRF models to a permutation baseline, for which the EEG data was paired with time-shifted speech signals over 100 permutations. We found an increase of prediction accuracy across age that was specific to the native language (mixed-effects model interaction: t(75) = 2.46, p = .016; native: t(76) = 3.54, p < .001, non-native: t(1.06) = 1.06, p = .29), indicating an increase in infants’ sensitivity to native phoneme features with age. Fitted confidence intervals across the native age trajectory suggest that native categorical processing of phoneme features significantly deviated from baseline from an age of 16 months onwards. Importantly, the developmental trajectory across individual features showed a significant correlation between a feature’s age of acquisition and its average duration in continuous speech (r(16) = .53, p = .014). This means that infants display categorical sensitivity earlier to those features that extend over longer stretches of speech. This effect remained significant after controlling for the overall frequency of occurrence of each feature. This means that even at equal exposure, infants’ sensitivity to a given feature is higher when it tends to extend in time. Our results show that the developmental trajectory of phonological acquisition is a function of feature duration in speech. Rather than individual phonemes, longer feature stretches in speech may ideally fit infants’ extended temporal receptive windows. The electrophysiological slowness of the infant brain might lead to an initial focus on slowly alternating phonological features, progressing to faster features when higher electrophysiology frequencies become available during ontogenesis. This suggests that infants use feature timing to accommodate initial electrophysiological constraints and bootstrap into their native phoneme inventory.
Topic Areas: Development, Phonology and Phonological Working Memory