Counting on Language: Morphological Transparency and Children's Acquisition of Number Words and Numerical Concepts

Abstract

This dissertation investigates whether and how the morphological transparency of number words in a child's language shapes the acquisition of number knowledge. Number words vary considerably across the world’s languages in how transparently they reflect the underlying Base-10 structure of the number system. In some languages, such as Mandarin and Tamil, number words are highly transparent - i.e., the word for ‘eleven’ literally translates as ‘ten-one,’ making the compositional logic of the counting system explicit. In others, such as English, Turkish, and Hindi, number words are partially or entirely opaque, requiring children to rely more heavily on rote memorization. This dissertation presents three empirical studies that examine how this cross-linguistic variation affects three key aspects of early numerical development: counting fluency, knowledge of the cardinal principle, and understanding of the Successor Function and how the count list implements it. Study 1 compared 363 children aged 3 to 6 acquiring Mandarin, English, and Turkish - three languages that differ systematically in the transparency of their number words. Study 2 compared 230 children aged 4 to 7 acquiring Hindi and Tamil - two languages spoken in India that differ dramatically in number word transparency while sharing comparable cultural, socioeconomic, and educational contexts, providing a uniquely controlled test of the causal role of number word transparency on number learning. In both studies, children acquiring languages with more transparent number words demonstrated higher counting fluency, as measured by a Free Counting task where we observed how high children could count without errors. Our analyses of the error patterns showed that counting failures were clustered systematically at morphologically difficult portions of the count list - particularly at decade boundaries in languages with opaque decades, and throughout the count list in Hindi, where even within-decade number words are morphologically non-transparent. Children acquiring transparent number systems performed better on the Unit Task, which served as a measure of children’s understanding of how the count list implements the Successor Function. As was the case in the Free Counting task, this advantage was most pronounced on trials requiring children to navigate decade boundaries - consistent with the hypothesis that transparent decade morphology specifically supports the discovery of productive morphological rules for number word composition. A third study examined English-speaking children’s productive morphological knowledge of number words using a novel number word paradigm, finding that 5-to-6-year-olds demonstrate some generalization of within-decade additive rules to unfamiliar lexical forms, though this knowledge remains incomplete and locally constrained. Together, these findings demonstrate that the morphological structure of number words is a consequential feature of the linguistic input children receive, affecting their learning of counting, as well as the development of foundational numerical concepts. Our findings thus contribute to ongoing debates about the role of language in cognitive development and have implications for early math education, particularly in contexts where the language of instruction has opaque number words.