Child speech is highly variable. The speech apparatus—the vocal tract, tongue, teeth, and vocal folds—develop at different rates for different children, which helps explain some of the variability in children’s speech. For example, the ratio of the oral to pharyngeal cavities changes as children age, making it difficult to establish reliable articulatory routines (Smith and Goffman 1998; Vorperian et al. 2005). While anatomy does play this undeniable role in child speech development, this dissertation focuses instead on components in the child’s environment that may explain their speech patterns. To do so, the studies here report on speech development in bilingual children acquiring South Bolivian Quechua (henceforth “Quechua”) and Spanish in a mid-size town in Bolivia. Chapters 2 and 3 examine the environmental effect of Quechua’s linguistic structure on speech development while chapter 4 examines the role of quantity of language exposure.
Chapter 2 examines how a phonological factor - vowel inventory—interacts with speech devel- opment. In Quechua, there are just three phonemic and two allophonic vowels. Chapter 2 asks if vowel inventory size mitigates acoustic variability in children aged four through ten. The study finds that children as young as four approximate adult-like acoustic targets, suggesting that child speech variability is contingent upon the language being learned. Still, children do not necessarily speak like adults. Using these vowel data, chapter 2 additionally finds that the children vary greatly in their ability to articulatorily compensate for their vocal tract morphologies, potentially explaining some of the large amounts of between-speaker variation that characterizes child speech.
Chapter 3 examines how another aspect of Quechua’s linguistic structure - its highly agglutinat- ing morphology - may interact with speech development. In Quechua, speakers construct words by supplementing root morphemes with a series of grammatical suffixes. Chapter 3 asks if this word composition could interact with children’s coarticulatory patterns. Here coarticulation is quantified using two novel acoustic measures that are less susceptible to the challenges that the child vocal anatomy poses for traditional spectral analysis. In experiment 2 of chapter 3, these measures are validated on a large corpus of four-year-old children acquiring English.
The central results of chapter 3 demonstrate that children and adults distinguish coarticulatorily between word environments: within morpheme (e.g. papa ‘potato’) and between morpheme (e.g. papa-pi ‘potato-loc). However, only children compensate for the morphologically complex words’ prosodic structure by shortening word duration. As a result, it remains unclear if the children’s spoken language patterns better reflect morphological or prosodic structure.
Finally, chapter 4 asks how children’s language exposure and use—in Quechua or Spanish— predicts the speech production outcomes from chapters and 2 and 3. In this study, each child’s bilingual language use patterns are computed from daylong audio recordings of the children’s lan- guage environments. Employing random sampling to annotate the recordings, chapter 4 efficiently estimates the children’s bilingual language environments: the annotation method required an aver- age of just 90 minutes of language category annotation from each recording to effectively estimate each child’s dual language exposure.
The chapter finds that children’s language exposure and use does indeed predict their speech patterns: children with monolingual Quechua mothers have tighter, less variable vowel categories than children with bilingual Quechua-Spanish or Quechua-dominant mothers. Additionally, children who use more Quechua throughout the day tend to distinguish more between the morphological environments tested in chapter 4. This last finding indicates that the more Quechua these children use, the better they are at analyzing and breaking down morphologically complex words.
Overall, the results from this dissertation demonstrate how myriad factors relating to linguistic structure and quantity of language exposure predict child speech variation. In doing so, this work also demonstrates how understudied languages—and novel methodological techniques like child-friendly acoustic measures and daylong audio recordings—can reveal aspects of children’s psycholinguistic representations, addressing long-standing questions in the field. Thus, this dissertation concludes that children face anatomical obstacles, such as an unstable oral to pharyngeal cavity ratio, that explain some of their speech variability. However, these anatomical factors co-exist with numerous elements of the children’s everyday linguistic environments to predict speech development.
