Learning Phonology from Surface Distributions, Considering Dutch and English Vowel Duration

Abstract

In learning language, children must discover how to interpret the linguistic significance of phonetic variation. On some accounts, receptive phonology is grounded in perceptual learning of phonetic categories from phonetic distributions drawn over the infant's sample of speech. On other accounts, receptive phonology is instead based on phonetic generalizations over the words in the lexicon. Tests of these hypotheses have been rare and indirect, usually making use of idealized estimates of phonetic variation. Here we evaluated these hypotheses, using as our test case English and Dutch toddlers' different interpretation of the lexical significance of vowel duration. Analysis of thousands of vowels of one Dutch and three English mothers' speech suggests that children's language-specific differences in interpretation of vowel duration are likely due to detection of lexically specific patterns, rather than bimodality in raw phonetic distributions.

A Cross‐linguistic Examination of Toddlers’ Interpretation of Vowel Duration

“…But the sparse evidence available suggests that in fact the phonologically long vowels can be realized with short durations in prosodically weak positions, arguing against this interpretation (Gussenhoven, ; Rietveld et al., ). Furthermore, Swingley (in press) found that in a sample of Dutch infant‐directed speech, the long vowels of Dutch were no less variable than the short vowels, and overall quite similar in their distribution to the durations of English long and short vowels. If, indeed, children learning English and Dutch hear similar vowel duration distributions in their daily lives, the explanation for children's language‐specific treatment of duration manipulations must lie elsewhere.…”

“…Perhaps they discover the durational features of their vowels by learning a prosodic model of their language in which multiple sources of variation control vowel duration, and in Dutch, a statistical difference then emerges that children can correctly attribute to intrinsic differences in vowel duration. Or perhaps specific words in children's experience exhibit distinctive duration features, and it is the lexicon that guides Dutch children to weighing duration more heavily (the case made by Swingley, in press).…”

“…One might expect that Dutch speech would exhibit bimodal durational distributions, that is, that measurement of a corpus of Dutch vowels would reveal a mass of short instances (the short vowels) and a separate mass of long instances (the long vowels). Instead, not only do the short and long Dutch vowels fail to form two modes, but also do not appear to be markedly more bimodal than English vowels are, even when considering infant‐directed speech (Swingley, in press). A similar observation has been made for Japanese infant‐directed speech, where there is no straightforward bimodal distribution of vowel duration, despite the fact that Japanese uses vowel duration phonemically (Bion, Miyazawa, Kikuchi, & Mazuka, ).…”

Do Infants Really Learn Phonetic Categories?

“…However, while some contrasts in laboratory speech are well-separated acoustically (Lisker & Abramson, 1964 ), categories overlap substantially in naturalistic speech, as in the second panel of Figure 2a (Antetomaso et al, 2017 ; Bard & Anderson, 1982 ; Bion et al, 2013 ; Hitczenko et al, 2020 ; Pollack & Pickett, 1963 ; Swingley, 2019 ). 4 Most models that have tested the feasibility of distributional learning for identifying phonetic categories have simplified the learning problem, for example, by using artificial data with low variability (McMurray et al, 2009 ; Pajak et al, 2013 ; Vallabha et al, 2007 ), focusing only on subsets of the categories infants would need to acquire (Adriaans & Swingley, 2017 ; de Boer & Kuhl, 2003 ; Gauthier et al, 2007 ), or limiting the training data to a single speaker (Miyazawa et al, 2010 ; Miyazawa et al, 2011 ).…”

Objectively measured teacher and preschooler vocalizations: Phonemic diversity is associated with language abilities

“…cal studies indicated that children selectively, and more readily, learn words that contain phonemes they are able to produce (Ferguson & Farwell, 1975;Leonard et al, 1981;Schwartz & Leonard, 1982;Stoel-Gammon & Cooper, 1984). In contrast, recent work employing computational models highlight the role of higher-level lexical knowledge in guiding children's acquisition of phonetic categories by providing informative cues about sounds that occur together in words (Feldman et al, 2013;Swingley, 2019;Swingley & Alarcon, 2018). Preschool children's perception of the phonological make-up of words, which is supported by the phonemic diversity of their own speech, is a strong predictor of long-range literacy outcomes, including reading skill (Wagner et al, 1997).…”