You are here: Home People Ingeborg Roete Research

Ingeborg Roete -


Description of PhD project:

During the first years of life children quickly learn the basics of their native language: phonemes, phonotactic patterns, words, phrase structure, etc. Many behavioral experiments assume a step-wise learning process: first children learn the sounds of the language, then they learn how to segment speech, followed by word meanings and more complex language features. Similarly, most computational models only investigate one aspect of language learning at a time. However, recent evidence from experimental studies on language acquisition suggests that infants can use information from (still incomplete) higher-level language representations to improve, for instance, their phonetic category learning.

This project aims to investigate whether (incomplete) knowledge of higher-level language representations can aid children in other, lower level, aspects of their native language. More specifically, I will look at whether infants can use phrase- and part word -level information to learn word meanings.

For this, I will mainly use a Bayesian computational modeling approach to incrementally build a model based on experimental findings from existing literature to mimics the infants’ learning process. First, I will start with a two-layer computational model (phrases and words) to see whether phrasal information can aid in learning word meanings. This model will be used to generate new behavioral predictions, which I can then test in experiments with infants. The experimental results can then be incorporated back into the model. Subsequently, the part word- level will be added to the model as a additional information source. This will result in a 3-layer model, which trades top-down and bottom-up information through three layers of representation during the learning process.

I strongly believe that child language acquisition can benefit from combining computational and experimental research methods. Computational models can be used to investigate the processes underlying the behavioral findings from experiments and help us to hypothesize about what is happening during first language acquisition.

Last checked 2017-08-18

Ingeborg Roete

Max Planck Institute for Psycholinguistics
PO Box 310
6500 AH Nijmegen
The Netherlands