Quantifying phonetic variation: A large-scale corpus analysis of coronal segments in English infant-directed speech

Abstract

Phonetic variation poses a challenge for language learners tasked with identifying the abstract sound categories (phonemes) of their target language(s). However, relatively little is known about the full extent of phonetic variability in naturalistic infant-directed speech (IDS), and how much of this variation can be explained by positional allophony. We phonetically annotated ∼31 000 tokens of coronal segments (/t/, /d/, /s/, /z/, and /n/) in naturalistic English IDS from the Providence Corpus (Demuth et al., 2006) in order to quantify the degree of phonetic variation present in the everyday speech directed to infants. We found that canonical variants are the most frequent variants for /d/, /s/, /z/, and /n/, but not for /t/. Additionally, every segment had more canonical instances in word-initial compared to word-final position. We are currently comparing the distribution of phonetic variants in specific phonological environments (e.g., tapping environments) in IDS with naturalistic adult-directed speech. We discuss the implications of these results for infants’ ability to identify the canonical (phonemic) variant from a set of phones, and uncover when and where positionally governed phonetic variants must surface.