Language is intrinsically polysemantic. The social-contextual tools that help us elucidate the ambiguity of each language utterance form what we call ‘common ground’, a safe territory of mutual agreement between the people taking part in the dialogue. How do we use this common ground though? Tracing the psychological foundation of language in children’s mechanisms of language acquisition can help us understand this process. The EU-funded ModelingCommonGround project merged developmental and computational approaches to focus on this topic. The research was undertaken with the support of the Marie Skłodowska-Curie programme.
Understanding how children acquire language
Children learn language in a rather complex environment that is abounding in dissimilar – and at times conflicting – information sources of what words mean. A common approach in research on children’s language learning is to isolate one of these information sources and test if children can use it. ModelingCommonGround emphasised a different rationale. “Our goal was to study multiple information sources at the same time. Furthermore, we wanted not just to see if children use multiple information sources, but also to specify how they do this,” says research fellow Manuel Bohn. To describe how children integrate different information sources during word learning, ModelingCommonGround used a computational cognitive model. In contrast to a verbal theory, such a model specifies the hypothetical computations that underlie word learning. Furthermore, it enables quantitative predictions about how children behave in new conditions. The project put this model in a dialogue with data from behavioural experiments. More specifically, it used experiments to measure children’s developing sensitivity to individual information sources and then used the model to predict what should happen if these separate sources were combined. The next step was to compare these predictions to new data from behavioural experiments in which multiple information sources were manipulated at one time. “We assume that there is a general way in which children integrate information during word learning. This process, we think, is stable and does not change over time. What we believe is changing is children’s sensitivity to each information source,” explains Bohn. “Instead of thinking of development as progressing through different stages or phases, we believe it is a continuous process that unfolds over time.”
Progress through trial and error
After the project model was tested with adults, the same experimental designs were used with children. However, they did not work. “Children did not find them interesting and showed no signs of understanding what was going on. It took us a couple of months to implement the basic experimental design in a child-friendly way,” notes Bohn. So far, the model has enabled predictions about a group of children. The next focus will be on studying individual differences through the same successful approach. “I hope that our findings will be seen as a model for studying the complexity of children’s learning environment. Since the beginning, we have made sure that we document and report everything we do transparently. Furthermore, all materials, data and scripts are publicly available. All our designs and analysis were preregistered. I hope that other researchers can use our work and build on it,” says Bohn.
ModelingCommonGround, children, information sources, common ground, language learning, model