Free Word Associations Correspond to Contiguities Between Words in Texts*

Abstract

A free associative response is the first word a person comes up with after perceiving another word, the so-called associative stimulus. People commonly associate hot to cold, church to priest, and hard to work. According to traditional association theory this behaviour is the result of learning by contiguity: ''Objects once experienced together tend to become associated in the imagination, so that when any one of them is thought of, the others are likely to be thought of also, in the same order of sequence or coexistence as before'' (James, 1890). This explanation has been rejected by cognitive psychologists who explain the production of associations as the result of symbolic processes which make use of complex semantic structures (Clark, 1970). We will show, however, that human associative responses can be predicted from contiguities between words in language use. This finding supports the hypothesis that the behaviour of participants in the free association task can be explained by associative learning.

How well can a corpus-derived co-occurrence network simulate human associative behavior?

“…The strength between two perceived events increases by a constant fraction of a maximally possible increment at each co-occurrence, and decreases in the opposite case. Wettler et al (2005) have shown that this mechanism can be replicated by looking at word co-occurrence frequencies in large text collections. But there had been earlier corpus-linguistic work: For example, Wettler & Rapp (1989) compared several association measures in order to find search terms to be used for queries in information retrieval.…”

“…After having lemmatized the data with the NLTK we obtained a list of 5910 test items which is considerably more than the usual 100 used in many previous studies (e.g. Wettler et al, 2005).…”

“…Unlike previous work (Wettler et al 2005;Church & Hanks, 1990) which is described in the terminology of the well known vector space model, in the construction of the current system we had a graph-based approach in mind so we describe the system in such terms. We built up a graph on the basis of the nouns, verbs, and adjectives occurring in the corpus, these tokens being the nodes of the graph.…”

“…Wettler et al, 2005;Tamir, 2005, Seidensticker, 2006. The main differences are that we categorize our stimulus words and present results for each class, and that we have a stronger focus on the graph aspect of our network.…”

Extracting Semantic Prototypes and Factual Information from a Large Scale Corpus Using Variable Size Window Topic Modelling

“…A comparison of human association norm and LSAmade association lists can be found in [4] and it should be the base of the study. Results of the other preliminary studies based on such a comparison: [5], [13], [14], show that the problem needs further investigation. It is worth noticing that all the types of research referred to, used a stimulus-response association strength to make a comparison.…”