Conventional thesauri and other word classification schemes used in information retrieval systems typically reflect the formal meanings of words. The central thesis of this paper is that associative meanings may be more relevant to nonspecific search strategies such as browsing.

The authors define the associative meaning of a word to be the distribution of response words to the given word in a word association experiment. Thus two words are similar with respect to their associative meanings if they have quantifiably similar distributions. Techniques for empirically determining these distributions, and quantifying word associations, are well established. The authors conducted such a study and built what they call a “term association thesaurus.” They then conducted a series of carefully controlled experiments to evaluate the thesaurus.

The study is well designed and the results should accurately reflect their underlying subject database and population. The research methods are appropriate, well documented, and generalizable. The results are interesting but inconclusive, since data from the experiments had not yet been analyzed.

The thesaurus is necessarily specific to the authors’ subject database and population and therefore will need to be updated as the database, population, and population needs change. The authors suggest that the information retrieval interface might elicit users’ associations during the query formulation process and thus constantly update the word association measures.

The paper is clear, direct, and well organized. It will be of interest to both practitioners and researchers concerned with the user interface to information retrieval systems.