A cross-linguistic analysis of the response distributions of restricted word associations

One-hundred Ss in each of 12 widely divergent linguistic communities were administered a standardized restricted word-association test consisting of 100 substantive stimuli. The Ss were instructed to provide a single response which conformed to the requirements of substitutability in a test frame designed to restrict responses to qualifiers only. The total frequency of all unique responses, excluding grammatically inflected responses, was tabulated. Categories of equal frequency of occurrence were determined and the distribution of the number of responses sharing the same frequency of occurrence was plotted. It was hypothesized that these distributions should substantially conform to a theoretical distribution of the lognormal form, since many aspects of the word-association task have high similarity to the generative rules of the lognormal distribution. The obtained distributions were found to conform sensibly to the hypothesized distribution. An analysis of the variance explained by the lognormal equations of best fit to the transformed points indicated that very little variance remained unaccounted for by the hypothesized distribution. Accordingly, variations of the estimated parameters were examined for clues as to the nature of the processes these parameters might reflect. The concept of stereotypy of response was introduced and defined as the degree of response uniformity across both Ss and stimuli. More generally, it was suggested that this stereotypy could be expected to be the result of stable linguistic conventions. The individual's responsiveness to these conventions was assumed to be a function of his sensitivity to response repetition within the limits established by the speech community.