Mixed Methods Ph.D. Researcher
Current Projects
Project | 01
Quantifying and Visualizing Variability in Infant Directed Speech
When we speak, we rarely pronounce words exactly how you might find them in a dictionary. This means that our speech is highly variable and
Although casual adult directed speech (ADS) contains a vast amount of phonetic variation, infant directed speech (IDS) has previously been thought to be more canonical (with less phonetic variation). Recent work has suggested that IDS and ADS are actually equally variable. Chong, Lin, and Sundara investigated the extent of phonetic variation in the Providence Corpus based on the position of coronal segments in the word, and found that in onset position, canonical variants were the most frequent variants; in medial position, the canonical variant was still most frequent though to a lesser extent; and in coda position, the canonical variant was not the most frequent variant for either /t/ or /d/, with large differences by segment. In my MA, I am extending this work to address the question: what is the level of phonetic variation in bilabial and velar stops and postalveolar fricatives in IDS? I then plan to design and run an experimental study, using segments extracted from the corpus, to determine whether infants prefer the environmentally appropriate or canonical variant of a segment, in order to pinpoint when the preference for the appropriate/correct form of a segment begins to arise over the course of an infant’s development – getting at the ultimate question: to what extent is phonetic variation in the input beneficial for language development?
Project | 02
Computational investigation of extralinguistic cognition in developmental parsing
Without explicit instruction, children learn language at a spectacular rate: by 5-6 years old, most children are able to speak and understand language quite well. As children learn their language(s), other cognitive systems, like memory and inhibition, are developing as well. Because language development relies on these additional cognitive systems, difficulties that stem from these developing systems may influence children's ability to learn language. This project uses computational tools to model the real-time process of sentence understanding to provide insights into how developing memory and inhibition may influence language understanding in children. This project serves as an important step toward understanding how developing cognitive systems can impact children's language learning, as well as how differences in these systems can influence children's later language performance in school.
This doctoral dissertation research focuses on children's understanding of "temporarily" ambiguous sentences as a window into the effect of systems like memory and inhibition on language learning and understanding. Like adults, children commit early to interpretations of sentences that they are hearing in real-time. Unlike adults, however, they appear to have difficulty revising those interpretations if additional, incompatible information arrives later in the sentence. This difficulty has been proposed to arise due to constraints on memory and/or immature inhibitory control. In this project, the researchers develop two computational models of children's sentence processing, one of memory limitations and one of inhibitory limitations, to test these proposals and probe the interaction of these cognitive systems with developing linguistic knowledge. By developing these models, this research provides a fuller understanding of the influence of memory and inhibition on language development and comprehension in children.
Project | 03
Event perception with four-place predicates
Most theories of how children learn verbs center around a mapping of a linguistic label onto a conceptual representation of an event. However, all of the previous literature has only tested the acquisition of simple transitive and intransitive verbs, which typically only label events with 1 or 2 participants. However, the visual working memory system in infants is reported to cap out at 3 items. This limit poses an interesting potential challenge for the acquisition of four-participant events, such as "trade", which involve two people trading two items. How might an infant learn a verb like 'trade' if their developing cognitive system place a limitation on the number of participants they could track?
Very little is currently known about the conceptual representation of social exchanges like tradings in humans. The representations that have been presumed as they pertain to word learning have stemmed largely from the number of arguments the predicates that label these event types take -- however, this argument structure knowledge is precisely what a child is trying to acquire when they learn a verb. To understand the types of tools infants might be using as they learn verbs, we must first take a step back and determine the conceptual representation of these event types. In this doctoral dissertation project, I am examining the conceptual representation of trading events in adults and preschool-aged children by presenting them with carefully manipulated videos of trading scenes.