From World to Word and Back: An Empirical and Philosophical Investigation on Syntactical Bootstrapping

This thesis has two connected components: an empirical project and a philosophical project

In the empirical project I present an experiment investigating syntactic bootstrapping as a noun learning mechanism. Previous studies have shown that infants as young as 6.5 months of age can perceive the causal relationship in a simple Michotte launching event. Yet it is not until the age of 2 year that children show evidence of reasoning with causality (Gopnik & Sobel, 2000; Gopnik, Sobel, Shulz & Glymour, 2001; Nazzi & Gopnik, 2003; Sobel, Tenenbaum & Gopnik, 2004). One reason for this lag is language. A conceptual understanding of causality requires understanding causal language, and the developmental trajectory of causal language in early childhood remains unclear. Studies have shown that toddlers at 15 months of age can match the transitive structure of the sentence to a causative event through a mechanism known as syntactic bootstrapping (Jin & Fisher, 2014). In the current experiments, we tested whether 12-month-olds and 20-month-olds can rely on the same mechanism to acquire the association between the subject of the sentence and the causal agent of the event. Due to COVID-19 outbreak, the data collection process was interrupted. Only preliminary data was presented in this paper, with discussion focused on the implications of different potential outcomes.

In the philosophical project I present a theoretical review on the foundations of infant looking time paradigms. In the past decades, the looking time measurement has been the backbone of developmental psychology. Most of our understanding of infant perceptual and cognitive development came from research using at least one of the three looking time paradigms: habituation, familiarization, and Violation of Expectation. However, the myriad claims supported by infant looking time paradigms suffer from what Richard Aslin called a “many-to-one” problem: there are many different postulated hypotheses, but only one measurement available (Aslin, 2007). The vast underdetermination between the evidence and the interpretation originates from the lack of attention devoted to the theoretical foundations of the looking time paradigms. In this paper, I surveyed and compared the four most prevalent theories in the field: The Comparator Theory, the Multifactor Model, the Object File Theory, and the Dual Process Theory. I analyzed each theory’s strengths and weaknesses in the context of experimental design and data interpretation. I also compared the four theories against each other to assess their explanatory scope. I arrived at the conclusion that none of the theories currently available is sufficient to justify the connection between the evidence and the interpretation. In the future, more systematic investigations are needed to construct a more precise, quantitative interpretation framework to guide empirical research.