Frequency Stability of Kent-Rosanoff Words

Abstract

Shifts in the frequencies of the Kent-Rosanoff words are described by applying a lognormal model to distributions resulting from the use of two frequency counts, the Thorndike-Lorge count and the Kucera-Francis count. The former's categorization limits consideration to 44 words which had frequencies in, or lower than, the A category (50–99 words per million) of T-L. The predominant shift is toward lower frequencies of occurrence in the latter count. The amount of shift toward lower frequencies increases as the Thorndike-Lorge frequency of the words decreases. The average shift is 14 words per million. The findings suggest a severe limitation on the generalizability of experimental results which have used the Thorndike-Lorge source to control the frequency characteristics of the Kent-Rosanoff words.