About

Word Association

Word association data has been used to support the scientific study of language since the nineteenth century. Words produced spontaneously in response to cues can reveal important differences in the way individuals process language, and construe meanings. Judicious analysis of such “word association” data can help us to understand the ways in which words are found, shared and lost:

Word finding can refer to the acquisition of new words in our first language, and to the acquisition of new languages. In both cases, newly acquired vocabulary is integrated into an existing ‘mental lexicon’, and develops connections with other words and concepts. Word finding can also refer to lexical retrieval; the capacity to find and produce a word in fluent discourse. We are typically only aware of this search process when it is problematic, for example when we are temporarily unable to retrieve a word that we know we know. Word association data can help us to model the mental lexicon as a network, so that we can examine anchor points for newly acquired words, and routes for word retrieval.

Word sharing refers to the everyday phenomenon by which we share our understandings of words we use when we communicate with each other. Word association data can tell us about the limitations and boundaries of those apparently shared understandings, and the subtleties and nuances of meaning that may not be shared.

Word loss, as evidenced in cases of aphasia or dementia, presents as a temporary or permanent loss of the capacity to retrieve the meaning and/or form of words. Word association tasks can be an efficient and low-impact way to collect data in such cases. The resulting data can reveal information about the kinds of words that are inaccessible, the kinds of lexical connections that seem to be damaged, and the possible circumlocution routes that might aid communication.

Word association tasks typically present a participant with a single word cue, to which they must respond with the first word that comes to mind. This is repeated for a series of cues. In some cases, participants are asked to give more than one response. The resulting data set will usually consist of a set of cue words, with the responses to each word from a set of participants.

Approaches to the examination and analysis of word association data typically take one of two routes, which can be summarised as analysis of stereotypy and analysis of association type.  The first of these uses ‘norms lists’ from groups of respondents as reference data against which to measure the stereotypy or idiosyncrasy of responses from a target group or individual. The second uses taxonomies of response types to categorise the kind of connection between cue and response. Taxonomies vary greatly in terms of type and number of categories, and depend on the research question being asked of the data, but most include, for example, categories for words with similar meanings, and for words that have a collocational relationship.  

Despite its long tradition and some promising findings, evidence from word association research has been frustratingly inconsistent. This is partly because of the many influences on any individual’s association behaviour, but also because methodological approaches have been unsystematic and un-coordinated. By providing, through this website, stable datasets with distinct characteristics, and tested analytic approaches, we aim to facilitate further well-informed and robust research, which can realise the full potential of word association approaches.

We are grateful for support from the Arts and Humanities Research Council (AHRC), part of UK Research and Innovation (UKRI), and from Swansea University, in the creation of these resources.

The Team

Tess Fitzpatrick

Steve Morris

Theodore Mills