Contributions of pitch and spectral information to cortical vowel categorization. (March 2020)

• Record Type: Journal Article
• Authors: Frank, Marina; Muhlack, Beeke; Zebe, Franka; Scharinger, Mathias
• Is Part Of: Journal of phonetics. Volume 79(2020)
• Publication Date: 2020-03
• Subjects: Speech sound processing -- Electroencephalography (EEG) -- N1 -- Pitch -- F1 -- F2 -- Vowels; Phonetics -- Periodicals; Phonetics -- Periodicals; Phonétique -- Périodiques; Phonetics; Periodicals; Electronic journals; 414.05
• DOI: 10.1016/j.wocn.2020.100963
• Languages: English
• ISSNs: 0095-4470

Abstract

Highlights: Spectral and periodicity properties of vowels are investigated using EEG. The component of interest (N1) is sensitive to both properties in complex sounds. We compared matched speech and non-speech based on f0 and F1 frequencies. In a matched setup, N1 properties were influenced by f0 only to a small degree. Early vowel processing is dominated by spectral information. Abstract: The perception of vowels such as [a] or [i] is based on their spectral properties, i.e. resonance frequencies (formants), and on fundamental frequency (f0). Yet, it is unclear whether early neural indices of vowel processing are predominantly driven by f0 or by formant frequencies. A candidate neural index of early acoustic processing is the N1, a negative evoked potential of the human electroencephalogram (EEG), peaking between 80 and 150 ms after stimulus onset. The N1 has been found to reflect pitch differences in simple tones, but also general vowel differences. However, the relative contributions of formant frequencies and f0 to N1 amplitudes and latencies are not entirely clear. To this end, an EEG study on German vowels differing in both first (F1) and second (F2) formant as well as f0 (independently modified) was conducted. Results showed that f0 only influenced N1 latencies of vowels with low F1, while F1 influenced amplitudes and latencies for all vowels. The distance between F1 and F2 as well as F2 significantly correlated with N1 amplitudes, but not with N1 latencies. These results imply that early neural processing of vowels relies on higher spectral components (formants) rather than f0.