Needless to say, the Covid-19 lockdowns in the UK have been a substantial impediment to this work, so it's really good to see it finally complete.
Category production is a task where participants name as many members of a given category that they can think of within a time limit. By looking at the order and timing of the responses that people make, it can be used to probe the organisation of the human conceptual system. It can also be used as a diagnostic tool when this conceptual system becomes affected, for example by dementia.
In this report, we conducted a large, pre-registered category production experiment, and analysed the results. We found that the order and timing of the responses was explained both by the similarity of sensorimotor experience between the category label and named members, but also the linguistic proximity — that is to say, an association between words learned from their natural occurrences in everyday language.
Following this, we deployed a large-scale computational cognitive model, approximating the full breadth of an adult's conceptual system. The model tracked associations spreading from the category to sequences of potential members, using both linguistic and sensorimotor proximity. The results showed that indirect associations (e.g. thinking of fish as an example of animal because you first thought of cat) are crucial to explain the responses that people make. The best version of the model was able to explain the variables in the participant data as well as any typical human's responses. The model even reproduced some of the same errors that the participants made, like naming pineapple and kiwi as an examples of citrus fruit.
A preprint of the work is available on PsyArXiv.
Edit: it got published!