2018 Sat Session A 1100

Saturday, November 3, 2018 | Session A, East Balcony | 11am

The lexical advantage: Four-year-old children acquire words, not sounds
M. Cychosz, M. Erskine, J. Edwards

A prominent debate in phonological development concerns the representation of children’s early speech sounds. Much of the perceptual narrowing literature suggests that children acquire individual phonemes over the first 12 months of life (Kuhl et al. 2005). These phones are later strung together to construct words and build a vocabulary (Berent 2013; Tesar & Smolensky 1998). An alternative account suggests that children build a phonology from holistic chunks of language such as words (Metsala & Walley 1998; Vihman 2017) or multi-word phrases (Arnon & Christianson 2017). Over time, these chunks individuate, abstracting to resemble an adult-like phonology.

This paper presents the results of an experimental task that evaluates these alternative hypotheses. We predicted that children would be more accurate at producing a CV sequence in a real word (e.g. tiger) than a corresponding nonword (e.g. taiblor). If so, this LEXICAL ADVANTAGE could be evidence that children do not learn individual sounds. Instead, they acquire phonology via language chunks such as words (Gathercole et al. 1991). Furthermore, in line with previous work (Author et al. 2004), we hypothesized that the lexical advantage would be more pronounced for children with smaller expressive vocabulary sizes. This is because these children have less language over which they can generalize to construct a phonology.

N=102 children aged 3;3-4;4 (mean=3;8) completed two tasks: a real word lexical repetition task and a corresponding nonword repetition task. Each of the N=23 real words had a paired CV syllable in the nonword (e.g. tiger, taiblor) that occurred word-initially. Since the CV sequence was constant between the real word and the nonword, articulatory difficulty with the sequence was controlled and a lexicality effect could be isolated. Children also completed a battery of assessments including standardized measures of expressive and receptive vocabulary.

A trained phonetician perceptually evaluated participants’ repetitions as [+/- correct] according to a feature-based scale: place, manner, and voicing for consonants and length, height, and backness for vowels. The prosodic structure of each CV sequence was also binarily evaluated: if the child produced the minimum number of syllables, if the consonant was in onset, and if the vowel was in nucleus. These components constituted each CV accuracy score. A second transcriber verified a subset (10%) of the transcriptions.

A mixed-effects linear regression model was fit to the accuracy scores. Best model fit included random intercepts for the CV_sequence and Child. A significant effect of Word_Type (real word versus nonword) indicated that children were more accurate at producing the same CV sequence when it occurred in a real word than a nonword (β=2.48, ***); a significant effect of Vocabulary_Size indicated that accuracy increased as expressive vocabulary size increased (β=.03, ***) (see Figure). A significant interaction of Word_Type and Vocabulary_Size indicated that the lexical advantage was more pronounced for children with smaller vocabularies. This reinforces more than a decade of literature suggesting that phonology emerges from connections children make between items in their vocabulary: the more words a child knows, the greater their level of phonological abstraction.