What is free association and what does it measure?

Abstract

This paper reports the results of a study of free association in which participants were asked to produce the first two words to come to mind. The findings were used to estimate the reliability of indices of strength and set size for different types of items and to model free association as a retrieval task. When confined to first responses, reliability was generally high for both indices, particularly for words with smaller sets of associates and stronger primaries. When second responses were included, reliability declined. A second response added new but weak items to the set, and, when the primary associate was not produced on the first opportunity, it tended not to be produced on the second. Relative to when multiple responses are requested, first-response free association provides more reliable indices of the relative strength and set size for a word's strongest associates. A model of free association assuming that a strength distribution underlies each response provided a good fit to the data.Memory tasks and their theoretical implementations model everyday memory tasks. Research on free recall, cued recall, and recognition, as well as work with implicit memory procedures, capture some, ifnot the majority, of the characteristics of the memory requirements needed in interactions with the environment. In this context, it is interesting that a memory task that has been used in research for more than 100 years (Galton, 1880) has not attracted much theoretical attention. We refer to free association, a task that requires participants to produce the first word to come to mind that is related in a specified way to a presented cue (e.g., meaning, rhyme, makes a word). This task is used in everyday activities as a means for "collecting thoughts." For example, when one uses the Yellow Pages, free associating to a needed product or service can be helpful in determining an effective search heading. The same advantage is apparent when attempting to find information on the Internet or just the right word in writing and speech. Similarly, we are often called on to provide information in response to broad and unanticipated questions, such as "Why are you a Democrat?" In such cases, recovering the needed information can be more akin to free association to the question than to directed recall.Free association is a utilitarian production task that has been directed to a variety ofdifferent types of information, such as

Verb similarity: Comparing corpus and psycholinguistic data

“…Moreover, we experimented with three versions of word associations: 1) similarities calculated using only the first response of each participant; 2) similarities obtained using the first three responses at most, as in De Deyne et al (2013); and 3) all the responses obtained. We observed that taking into account the first three and all responses yielded higher correlations with corpus data than taking only the first response, in line with the findings of Nelson et al (2000) and De Deyne et al (2013). Overall, higher correlations were obtained when taking into account just the first three responses.…”

“…It is generally acknowledged that these relations have a broad semantic basis (Nelson et al 2000;Roediger et al 2001;McRae et al 2012;Brainerd et al 2008). Evidence suggests that we might find typical relations such as hyponymy, synonymy and antonymy (Clark 1970), although other types of relations such as cause-effect or part-whole (Hernández et al 2014) and thematic relationships (Peirsman and Geeraerts 2009) are more frequent.…”

“…We chose to collect several responses to characterize each stimulus, as not limiting the responses to the first that came to mind has shown higher correlation with human similarity ratings (Steyvers et al 2004) and it has proven to perform better in lexical access tasks (Chumbley and Balota 1984). Besides, Nelson et al (2000) and De Deyne et al (2013) suggest that these nonprimary responses contain useful information. In addition to these advantages, from the point of view of similarity calculation between verb senses, having more than one response per stimulus and participant helps in reducing data sparsity (De Deyne et al 2013).…”

Efeitos de priming semântico em tarefa de decisão lexical em crianças de 3ª série

“…Esta análise foi realizada após excluírem-se as respostas do tipo "não sei". A força de associação corresponde à maior probabilidade de uma resposta ocorrer em detrimento de outras, dada uma palavra-alvo específica (Nelson, McEvoy, & Dennis, 2000). A força de associação das palavras da lista variou entre 6 e 90 %.…”

“…Na análise do tamanho do conjunto de associados semânticos diferentes gerados para cada palavra-alvo, quando considerado o tamanho total do conjunto houve uma (Nelson et al, 2000;Nelson & Schreiber, 1992;Van Erven & Janczura, 2004). A análise de correlações entre os fatores mostrou que quanto maior a força de associação semântica entre as palavras menor é o tamanho do conjunto gerado, total e significativo.…”