Figure 1: An example for a phonological network based on the words a specific child knows at a specific age. The connections between the words exist when the Levenshtein distance between two words is smaller or equal .25.
The Development of Early Phonological Networks
Judith Kalinowski, Laura Hansel, Michaela Vystrčilová, Alexander Ecker, and Nivedita Mani
The influence of phonological similarity between words on the development of early vocabulary has caught the interest of language learning research in recent years. In order to investigate the influence of already learned words on the acquisition of new and phonologically similar words, the vocabulary of individual children must be analyzed over time, as each child learns in a different way. Previous research has used networks to represent children's vocabulary. However, these are based on data averaged over many children (Fourtassi et al., 2020; Siew & Vitevitch, 2020). Laing (2022) used data from individual children in her analysis, but only had a small data set. Because of the different data and the different phonological distances used in the previous studies, the results of the analyses differ: Fourtassi et al. (2020) showed that words that have many phonologically similar words in the children's linguistic environment are more likely to be learned earlier than words that have hardly any similar words in the environment (EXT). Laing (2022) found that previously learned words that are phonologically similar to many other previously learned words attract similar new words. Consequently, children would learn these first (INT). However, Siew & Vitevitch (2020) found that both INT and EXT are statistically significant predictors of word learning, but also argued that INT has a greater impact on phonological network growth. In the present work, we address the problem regarding different (non-longitudinal) data and phonological distances by using a large longitudinal data set of 1,565 Norwegian children, comparing three different phonological distances, and applying them to our model. This allows us to investigate whether early lexicons grow in an INT- or EXT-like manner and whether different phonological distances lead to different results.
Find the preregistration to this project on the Open Science Framework: https://osf.io/kd95f/registrations.
The code can be accessed via GitHub: https://github.com/JudithKalinowski/LongPhon.
Figure 2: Exemplary objects of two different categories which will be used in the experiment.
Form-Meaning Systematicity in
Early Referent Selection
Early Referent Selection
Judith Kalinowski, Ming Yean Sia, Janna Leismann, and Nivedita Mani
The preregistration to this project will be uploaded to the Open Science Framework.
The code will be made accessible via GitHub.