Over the summer I spent six weeks on a LuCiD summer internship supervised by Dr Katie Twomey and Dr Alissa Ferry, assisting with research into children’s acquisition of categories at the basic and global level, e.g. dog (basic) vs animal (global). I was tasked with a specific project which had multiple aims.
First we wanted to determine the relative frequency of basic and global level words in child-directed speech. Using this information we then planned to explore the link between the frequency of basic and global level words in the input and the age at which children learn and begin to use such words.
To do this, I created a list of basic and global level nouns based on the UK Communicative Development Inventory (UK-CDI; Alcock, Meints & Rowland. 2017). For each word, we extracted total frequency counts from UK corpora of child-directed speech on the CHILDES database (MacWhinney, 2000) using the CLAN database query software. We noted the age at which children learn to understand and produces each word based on CDI norms. Since norms regarding age of acquisition and production are not yet available for the UK-CDI, such norms were taken from the MacArthur Bates CDI for use in this project. Finally, we coded whether each word belonged to a basic or global category.
We were interested in whether a word’s frequency and category membership (basic, global) predited the age at which children learned to produce that word. We therefore ran multiple regression analyses of this data using the statistical software R to find the best fitting model.
We found no effect of whether a word belonged to a global or basic category on age of acquisition. Interestingly, however, the CDI items included very few global level words. Thus, this lack of effect could have been due to low statistical power for the global items; alternatively, global and basic level terms could be equally easy to learn. However, we did find a small but significant association between frequency and age of production, that is, the more a category label was used in the input the earlier the child would produce this label. The same was not found for age of comprehension as many data points clustered around the same age.
What does this mean?
We already know that a high-quality, language-rich environment is important for children’s language learning but this project adds to the existing literature with regard to the importance of quantity, i.e. the more commonly a category label is directed towards a child, the earlier the child will produce that label.
During the course of the internship I also attended a two-day LuCiD mini-conference in Liverpool where I was able to network with other researchers and listen to presentations by LuCiD researchers on their current projects. This gave me the opportunity to apply my analytical skills by asking questions about research methodologies as well as build on my existing knowledge of language processing models and connectionist modelling developed through my BSc Speech and Language Therapy degree so far. It also gave me plenty of food for thought for my final year extended project.
Attending fortnightly reading groups also gave me the chance to engage with recent literature in a lively and challenging way, while assisting with other LuCiD researchers’ experiments gave me an appreciation of the challenges of experimental design, all of which will be invaluable for me going forwards into final year. I also benefitted from being trained in how to use modern and mobile eye-gaze technology which will serve the members of the team well in their work with children in schools and in other settings outside of the lab itself.
Before I started the internship I was keen to see what a day in the life of a researcher might look like and whether a research career might be something I would consider. My passion has certainly not faded and seeing the variety that this kind of career can bring – no two days are ever the same – I look forward to seeing where my research interests take me going forwards.