Learning how to inflect verbs in different languages
Authors: Anna Theakston and Ben Ambridge, Division of Human Communication, Development and Hearing.
English has a very limited range of inflections that we add to the ends of nouns (dog/dogs) and verbs (jump/jumps/jumped) to convey different meanings, but many other languages have far more.
We know little about the ways in which children learn how to use the full set of inflections across a wide range of contexts.
In this project, we looked at three highly inflected languages (Polish, Finnish & Estonian). We studied how 2-5 year old children used different inflections by running large-scale elicited production studies, where native language learners were shown animations or pictures of different characters performing various actions, and asked to describe what they saw. We also built a computational model to test theories of how inflections are acquired using the distributional statistics of the language input and to compare to the child data.
We analysed the data using mixed effects models (frequentist and Bayesian). We found that for both nouns and verbs across languages, children made more errors with forms that have lower frequencies in the input, and they were also influenced by how many other words could be inflected in a similar way (e.g. sing/sang, ring/rang vs. bring/brought). These findings suggest that successful models of children’s acquisition of morphology need to be sensitive to the statistical properties of children’s input.
Applying open research practices
Both studies (Nouns and Verbs) exemplify high adherence to open science principles. The accompanying OSF pages include open materials, open data, and open code (in the form of scripts for analysis written in R) for both studies. The second study (Nouns) was fully pre-registered on the OSF in advance of data collection. Both studies are now open access publications. In addition, the data are also logged with the UK Data Archive providing a second route to access the study protocols, design, materials, data and analysis (in accordance with funder requirements at the time). Spin-off studies using our open materials are promoting replication.
Working across three languages posed challenges as we attempted to derive concrete, theoretically motivated predictions, establish how to implement our various frequency/PND measures and design appropriate child-friendly materials to enable comparison across languages. We were able to address these issues successfully by working with native speakers of the three languages as co-investigators to uncover and address many practical challenges and establish how best to rigorously test different theoretical accounts. This was possible because we had assembled a large research team with complementary expertise in managing large research projects and teams, child-friendly experimental design, computational modelling plus the necessary linguistic expertise.
Benefits of using these open research practices
Our use of open science practices in the form of open materials, data and code brought a high level of organisation to our large research team. This was essential to consistently labelling files and creating code that can be interpreted both within the team and externally. Our creation of open materials applicable to a range of languages has led to a number of spin-off studies, conducted by researchers looking at other languages. These kinds of replications are important for theory testing and development within our field.
Start early – the sooner you begin to plan the details of your study and start the pre-registration process, the clearer your thinking and the better your organisation will be.
Find out more
How the input shapes the acquisition of verb morphology: Elicited production and computational modelling in two languages
Language-general and language-specific phenomena in the acquisition of inflectional noun morphology: A study of Polish, Finnish and Estonian